MEASURING POLARIZATION

TECHNICAL FIELD

The invention relates to a method for measuring cloud-base distance and to an apparatus for measuring polarization.

PRIOR ART

The rate of cloud cover of the sky, being one of the most significant input parameters of climate modeling, is mostly determined from satellite images. In many cases, however, a need arises to estimate the rate of cloud cover at a given geographic location from the Earth's surface. In addition to cloud cover, information relating to cloud-base distance is also important in many fields of meteorology, especially in weather forecasting for airports.

In a number of meteorological stations, still nowadays, the rate of cloud cover is determined by visual estimation yielding generally rather inaccurate results. Meteorological devices capable, within certain limitations, of detecting the cloud pattern of the sky provide solution to this problem. The most state-of-the-art meteorological stations acquire data about cloud cover by means of computer analysis of color sky images taken by digital cameras fitted with 180° field-of-view lens, e.g. so-called fisheye-lens, mapping the entire sky.

One disadvantage of prior art cloud detectors is that they are not portable, as they require considerable power and computer connection, rendering them inadequate for on-site measurements. Such cloud detectors mostly perform cloud detection by means of a photometry-based algorithm, i.e. clouds are being separated from clear sky on the basis of intensity and color of the skylight. However, in addition to its wavelength (color) and intensity (brightness), light is also characterized by the direction of its polarization (direction of oscillation plane) and the degree of its polarization (proportion of light with characteristic oscillation plane). Polarized skylight has a characteristic polarization pattern, being dependent on the position of the Sun, the aerosol content of the air, as well as the presence of clouds and the height of cloud-base. Reliability of the method for cloud detection can be improved in a known manner by using the additional information of the degree of polarization and the direction of polarization in addition to the intensity distribution of skylight measured in the red, green and blue spectral ranges, which is the color photo of the sky. Using conventional photographic techniques only the wavelength and intensity of skylight can be measured, for the measurement of the polarization pattern of the sky an apparatus capable of measuring polarization is required, which, by way of example, may be a camera fitted with appropriate polarizing filters, and by which, using an optical device having appropriate field of view, the polarization pattern of the entire sky can be measured. There are numerous methods available in the literature for measuring cloud-base distance and there are also known apparatuses capable of measuring cloud cover.

In DE 10 2007 039 095 A1 and DE 43 02 622 A1 cloud cover measurements without using polarization information are disclosed.

In a study (J. A. North and M. J. Duggin, Stokes vector imaging of the polarized sky-dome, Appl. Opt, Vol. 36, pp. 723-730 (1997)) an apparatus for the measurement of polarization is disclosed, which examines the polarization parameters of the skylight by means of a four-lens camera using linear polarizing filters with different transmission directions.

In two other studies (K. J. Voss and Y. Liu, Polarized radiance distribution measurements of skylight. I. System description and characterization, Appl. Opt, Vol. 36, pp. 6083-6094 (1997); Y. Liu and K. J. Voss, Polarized radiance distribution measurements of skylight. II. Experiment and data, Appl. Opt, Vol. 36, pp. 8753-8764 (1997)) an apparatus for measuring polarization is disclosed comprising an exchangeable linear polarizing filter, a 180° field-of-view lens and a CCD camera. This apparatus is capable of determining the polarization pattern of the entire sky by means of computer analysis of three images of the sky recorded with three different orientations of the transmission direction of the polarizing filter.

A portable, 80° field-of-view apparatus using fisheye lens camera equipped with three linear polarizing filters having different orientation of the transmission directions and being capable of measuring polarization is disclosed in two studies (J. Gal, G. Horvath, V. B. Meyer-Rochow and R. Wehner, Polarization patterns on the summer sky and its neutral points measured by full-sky imaging polarimetry in Finnish Lapland north of the Arctic Circle, Proc. R. Soc. Lond. Ser. A, Vol 457, pp. 1385-1399 (2001); I. Pomozi, G. Horvath and R. Wehner, How the clear-sky angle of polarization pattern continues underneath clouds: full-sky measurements and implications for animal orientation, J. Exp. Biol., Vol. 204, pp. 2933-2942 (2001)).

The disadvantage of all aforementioned apparatuses for measuring polarization is that in case of rapid changes in the sky, such as rapid movement of a cloud, they cannot be appropriately used, since the three images of the sky are not recorded simultaneously, but are captured successively with a few seconds of delay due to the single wide-angle lens and detector. As the clouds move in the meantime, the time distribution of the images will cause distortions to polarimetry.

In other studies (Barta A., Horvath G., Gal J., Suhai B. and Haiman O., Cloud detection from the Earth by means of 180° field-of-view imaging polarimetry, Fizikai Szemle, Vol. LI, No. 10, 315-319 (2001); Barta A., Horvath G., Gal J., Suhai B. and Haiman O., Cloud detection from the Earth by means of 180° field-of-view imaging polarimetry, Fizikai Szemle, Vol. LI, No. 1 1 , 355-362 (2001); G. Horvath, A. Barta, J. Gal, B. Suhai and O. Haiman, Ground-based full-sky imaging polarimetry of rapidly changing skies and its use for polarimetric cloud detection, Applied Optics, Vol. 41., No. 3., pp. 543-558 (2002)) a three-lens, three-camera, 180° field-of-view apparatus for the measurement of polarization is disclosed, which is capable of taking the three images of different linear polarization simultaneously. A disadvantage of this apparatus, however, is that the three images are taken from positions a little apart from one another, therefore, the respective fields-of-view of the cameras are not identical.

In US 201 1/0050854 A1 an instrument equipped with a fish-eye lens for measuring cloud cover is disclosed, the instrument also making use of polarization information.

There are other studies as well, engaged in the examination of polarization characteristics of clouds (A. J. LePage, A. T. Stair, Jr., R. J. Jordano, P. C. Joss, J. Devore, J. A. Kristl, B. P. Sanford and J. H. Summers, IR/Visible polarization measurements of scattered solar radiation from clouds, Aerospace Conference, 1999, IEEE Proceedings, Vol 4., pp. 239-257 (1999); J. G. DeVore, J. H. Thompson, R. J. Jordano, A. T. Stair, C. Trowbridge, P. C. Joss, S. A. Rappaport, J. KristI and R. A. McClatchey, Effect of cloud polarization properties on target discrimination, Aerospace Conference, 2001, IEEE Proceedings, Vol 4., pp. 1985- 1994 (2001); O. Stahli, C. Matzler, A. Murk, N. Kampfer, Sky measurements with the imaging polarimeter SPIRA at 91 GHz, Microwave Radiometry and Remote Sensing of the Environment (MicroRad), 2010, 1 1th Specialist Meeting on Digital Object Identifier, pp. 181-186 (2010)). In addition to analyzing polarization pattern of the sky, there are numerous other solutions for determining the distance of clouds from Earth's surface.

A widely-used method for the determination of cloud-base height is the laser- based LIDAR (Light Detection And Ranging) technology. A disadvantage of the LIDAR measurements, however, is that it is capable of measuring the absolute distance of cloud-base in one direction at a given time, i.e. in only one point of the sky-dome.

A study (A. J. Poyer and R. Lewis, Adapting the micropulse Lidar for use as a reference for cloud measurement, Fourth Symposium on Lidar Atmospheric Applications, (2009)) discloses a cloud-base height measurement with LIDAR. It discusses the polarization characteristics of clouds, as well.

In US 4,355,893 a laser-based method for measuring the cloud-base height is disclosed. In US 4,605,710 a measurement of the cloud-base height by means of an optical system is disclosed. In US 6,035,710 infrared radiation for determining the base height of clouds is used. In US 2001/0043323 A1 a method for determining the spatial coordinates of a cloud is disclosed. According to ES 8800752 A1 the base height of a cloud is determined by means of an optical system.

There are various methods for measuring cloud-base height presented in US 6,281 ,969 B1 and in other studies (I. Genkova, G. Seiz, P. Zuidema, G. Zhao and L. Di Girolamo, Cloud top height comparisons from ASTER, MISR, and MODIS for trade wind cumuli, Remote Sensing of Environment, Vol. 107, pp. 211-222 (2007); A. A. Kokhanovsky, C. M. Naud and A. Devasthale, Intercomparison of ground- based radar and satellite cloud-top height retrievals for overcast single-layered cloud fields, IEEE Transactions on geoscience and remote sensing, Vol 47., No. 7, pp. 1901-1908 (2009); E. Brocard, M. Schneebeli and C. Matzler, Deriving winds at cloud-base height with an infrared camera, IEEE Transactions on geoscience and remote sensing, Vol 47., No. 10, pp. 3319-3325 (2009); F. M. Janeiro, P. M. Ramos, F. Wagner and A. M. Silva, Developments of low-cost procedure to estimate cloud-base height based on a digital camera, Measurement, Vol. 43, pp. 684-689 (2010)).

Other studies present cloud detection made from above (from satellites or aircrafts) (F. Lu, J. Xu, W. P. Menzel, C. S. Velden, Geometric Cloud Top Height assignment by geosynchronous meteorological satellite images, Geoscience and Remote Sensing Symposium, 2009 IEEE International, IGARSS 2009, Vol. 3, pp. III-605 - III-608 (2009); D. Harding, P. Dabney, J. Abshire, T. Huss, G. Jodor, R. Machan, J. Marzouk, K. Rush, A. Seas, C. Shuman, X. Sun, S. Valett, A. Vasilyev, A. Yu and Y. Zheng, The slope imaging multi-polarization photon-counting Lidar: an advanced technology airborne Laser altimeter, NASA Earth Science Technology Forum, June 22-24, 2010; O. Hagolle, A. Guerry, L. Cunin, B. Millet, J. Perbos, J.-M. Laherrere, T. Bret-Dibat, POLDER level 1 processing algorithms, Proceedings of SPIE Aerosense 96:"Algorithms for Multispectral and Hyperspectral Imagery II" Orlando, (1996), pp. 308 - 3 9).

A common disadvantage of the methods and apparatuses for determining the cloud-base height is that in most cases they are incapable of determining the height of cloud-base on the full sky portion visible from a given geographic location, and generally require complex apparatuses comprising, in many cases, expensive optical solutions.

In the light of the known solutions, there is a demand for developing a method enabling determination of cloud-base height/distance on the widest possible part of the sky visible from a given geographic location, at the same time requiring the least possible devices. There is also a demand for providing an apparatus for measuring polarization, preferably the polarization of sky-dome, being capable of taking three different linear polarization images simultaneously from one position, i.e. by the use of one image recording device, having essentially the same field-of- view. DESCRIPTION OF THE INVENTION

It is an object of the invention, therefore, to provide a method for measuring cloud- base distance and an apparatus for measuring polarization, which are free from the disadvantages of prior art solutions.

It is a further object of the invention to provide a method for measuring the cloud- base distance, wherein, in a nearly 180° field-of-view the relative distance of clouds in a sky-portion visible from a given geographic location, and their relative distance from Earth's surface can be measured by the use of imaging.

It is a further object of the invention to provide a method for measuring the cloud- base distance, wherein the absolute distance of a cloud located in one point of the examined sky-portion is calibrated, and the distance of all other clouds from the Earth's surface is determined geometrically from the relative distances to the calibration point.

It is a yet further object of the invention to provide an apparatus for measuring the polarization, which is capable of capturing linear polarization images of the sky by means of a single recording device through at least three linear polarizing filters of different transmission directions at the same time and essentially from the same position, i.e. having substantially identical fields-of-view, preferably a wide angle field-of-view and especially preferably up to 80° field-of-view.

The objects of the invention can be achieved by the method according to claim 1 and the apparatus according to claim 22.

BRIEF DESCRIPTION OF DRAWINGS

Preferred embodiments of the present invention are described below by way of example with reference to the following drawings, where

Fig. 1 is a diagram illustrating a curve fitting, being part of the method according to the present invention, Fig. 2 is a schematic representation of a stereo measurement, Fig. 3 is a perspective view of a preferred embodiment of the apparatus according to the invention capable of implementing the method according to the invention,

Fig. 4 is a perspective view of a beam splitter prism system being part of a preferred embodiment of the apparatus according to the invention, showing polarizing filters and an image recording device,

Fig. 5 is a perspective view of the light splitting prism of Fig. 4, from another point-of-view,

Fig. 6 is a perspective view of the polarization matrix being part of another preferred embodiment of the apparatus according to the invention, indicating a Bayer-filter and an image recording device, and

Fig. 7 is the perspective view of the apparatus according to prior art technology capable of implementing the method according to the invention. MODES FOR CARRYING OUT THE INVENTION

The method according to the invention can be implemented by means of an apparatus for measuring polarization, by which, according to the method of the invention, linear polarization images can be taken of a sky-portion by means of at least three, preferably three, linear polarizing filters with different transmission directions, from essentially the same point of view and essentially at the same time. Furthermore, points equivalent with each other can be selected on the linear polarization images by means of the apparatus, then the degree of linear polarization and total intensity of an incident light beam arriving on the points equivalent with each other can be determined using of the intensities measured at the equivalent points. With the apparatus, a relative cloud-base distance is determined for the points equivalent to each other based on the degree of linear polarization and the total intensity as well as on a pre-determined degree of linear polarization and a pre-determined total intensity belonging to a clear sky. The method according to the invention will be described henceforward, by way of example, on the basis of three linear polarization images taken by using three linear polarizing filters with different transmission directions. ln a preferred embodiment of the method according to the invention, at least three linear polarization images are taken, especially preferably in the three different visible spectral ranges (red, green, blue), by means of digital image recording devices. The digital images are taken by means of an image recording device consisting of light detection elements. The images may be taken by means of a color-sensitive image recording device, wherein the individual light detection elements capture color information. In this case, separate light detection elements are used on the digital image recording device for the at least one color, i.e. in each different spectral range in the points equivalent with each other, preferably for separately processing the components of the individual colors. The images can also be taken by such color-insensitive image recording device, which is made to be color-sensitive by means of an array of color filters, such as Bayer-filter. The Bayer-filter is an array of color filters, wherein the color filters, each equaling the size of a single light detection element, are organized in groups of four, and the square-formed, by way of example, red blue and green color filters are arranged as a matrix. Of the three elemental color filters (e.g. red, green and blue) there are groups of four elements organized by using two of one color, e.g. green, arranged, by way of example oppositely, in each group of four. The non-color-insensitive image recording device can be turned into a color-sensitive image recording device having lower resolution, one-fourth of the resolution in this example. By means of known algorithms, it can be achieved that the color-sensitive image recording device would have an optical resolution close to that of a color- insensitive image recording device comprising identical number of light detection elements arranged at the same density. However, as opposed to the above, the method according to the invention may be implemented by means of a color-insensitive image recording device or such an image recording device, which has not been made color-sensitive, e.g. an image recording device for taking black-and-white images, as the method according to the invention is principally based on the use of polarization information. In the present embodiment, therefore, points equivalent with each other are selected per spectral range, the degree of linear polarization and total intensity of the incident light beam arriving at the points equivalent with each other is determined per spectral range, and relative cloud-base distance is determined per spectral range based on the degree of linear polarization of the clear sky and on the intensity of the clear sky, and on the degree of linear polarization and on the total intensity. As a first step of the method according to the invention, linear polarization images of a sky-portion are taken by means of three linear polarizing filters with different transmission directions from essentially the same position and essentially at the same time in three different spectral ranges.

In the next step, points equivalent with each other are selected on the linear polarization images of the three spectral ranges, i.e. a search is carried out on each linear polarization image for the points equivalent with each other, i.e. those representing the same point in the sky (or in other words, corresponding to the same direction in the field of view) - in the case of a digital image recording device for the respective light detection elements, and in the case of a color- sensitive digital image recording device for the respective light detection elements with different spectral sensitivities.

Then, by means of the respective intensity values of the points equivalent with each other, in the three spectral ranges, the degree of linear polarization of the three colors and the total intensity of the three colors of the incident light arriving on the points are determined. To this end, the light intensity values are used, which are measured in the points equivalent with each other through the three polarizing filters with different transmission directions in the three different spectral ranges.

Fig. 1 illustrates the determination of polarization information, degree of polarization, intensity, of a given light detection element, i.e. a given point in a spectral range. For this reason, the intensities of light measured in the given spectral range through the three linear polarizing filters with different transmission directions are put on a graph, where the angle of the transmission axis of the polarizing filter relative to the vertical is set on the horizontal axis, and the relative intensity of light is set on the vertical axis. Accordingly, the highest value of intensity can be 1 (which means that the light can pass through the filter without any hindrance), while the lowest possible value is 0 (which means that the light is totally blocked by the filter). Fig. 1 shows an exemplary case for the illustration thereof, representing intensity values measured with polarizing filters, the transmission directions of which were oriented at 0°, 60° and 120° (indicated with X), in case of partially linearly polarized incident light. A sinusoid function is fitted to the three acquired points, with a period of 180°. Periodicity of 180° means that a linear polarizing filter affects the light in the same manner as another linear polarizing filter rotated by 180°. For fitting the curve, it is necessary to have the image taken by the at least three linear polarizing filters with different transmission directions, as the three-parameter sinusoid function can be fitted to three different points. The maximum point of the fitted function, i.e. the angle value on the horizontal axis where the fitted function has its maximum value, gives the angle of polarization of the incident light beam, i.e. its direction of polarization, because the direction of transmission of the polarizing filter is parallel to the direction of polarization of light. The intensity value taken at the maximum point can be less than one, because light may have polarized as well as unpolarized components. Therefore, even in a case when the direction of polarization of the analysed light beam is identical with the direction of transmission of the polarizing filter, the measured intensity may be less than one due to the unpolarized component. In addition to l _ma x maximum intensity, the fitted function gives an \ _m minimum intensity as well. The degree of linear polarization can be calculated from the maximum and minimum values of intensity by means of the following equation:

Ρ ⁼ ('max ^— lmin)/(lmax Imin)-

In the present embodiment of the method according to the invention in all three (red, green, blue) spectral ranges (color channels) three different transmitted intensities of light are obtained according to the three linear polarizing filters with different transmission directions. Accordingly, a value of the degree of polarization and a value of the direction of polarization are obtained in all three color channels in the present embodiment. The light of a clear, cloudless sky has a characteristic pattern of direction of polarization and degree of polarization, which is well represented by the Rayleigh model, and more precisely by the Berry model. The Rayleigh model (J. W. Strutt (Lord Rayleigh), On the light from the sky, its polarization and colour, Philosophical Magazine Series 4, Vol. 41 , Issue 271 , pp. 107-120 (1871); J. W. Strutt (Lord Rayleigh), On the light from the sky, its polarization and colour, Philosophical Magazine Series 4, Vol. 41 , Issue 273, pp. 274-279 (1871); K. L. Coulson, Polarization and Intensity of Light in the Atmosphere, A. Deepak Publishing, Hampton, Virginia, USA (1988) describes the spatial distribution (according to direction) of the intensity and polarization of sunlight singly scattered in the atmosphere on the imaginary celestial sphere. According to the model:

(i) the polarization of skylight is independent from the wavelength of light,

(ii) the maximum degree of polarization is at 90° from the direction of the Sun, from where it gradually decreases according to a simple mathematical formula to 0 by moving towards the Sun and anti-Sun,

(iii) the direction of polarization (plane of oscillation) of skylight is always perpendicular to the so-called scattering plane, which is defined by the ground-based observer, the Sun and the observed point in the sky.

According to the Rayleigh model, there are two so-called neutral (polarization- neutral) points in the sky, the Sun and anti-Sun, from where non-polarized light propagates. In fact, multiple scattering of the sunlight in the atmosphere is also possible, which would cause a more complex polarization distribution than the Rayleigh model. The two most significant differences compared to the Rayleigh model:

(i) existence of three neutral points observable from the Earth's surface, and their differences from the Sun and anti-Sun, and

(ii) significant decrease of the degree of polarization of skylight estimated by the Rayleigh model due to the depolarizing effects of multiple light scattering.

The Berry-model (M. V. Berry, M. R. Dennis, R. L. Lee Jr., Polarization singularities in the clear sky. New Journal of Physics, Vol. 6, 162 (2004)) describes the degree of polarization and the pattern of polarization direction of the cloudless sky with complex functions more precisely than the Rayleigh-model, and also provides the positions of the neutral points. The mathematical descriptive functions of the Berry-model have proper physical reasons, traceable to multiple scattering (J. H. Hannay, Polarization of sky light from a canopy atmosphere, New Journal of Physics, Vol. 6, 197 (2004)). If the pattern (i.e. distribution in the sky according to direction) of the degree of polarization and direction of polarization determined by the apparatus for measuring the polarization of skylight, i.e. of the degree of polarization and direction of polarization received in the aforementioned manner, is compared point-by-point to the theoretical patterns estimated by the Rayleigh-model or Berry-model, then the presence of the clouds can be deduced from the differences from the theoretically calculated patterns: the higher the differences, the more probable is the presence of a cloud in a given point of the sky. If it is not the probability of cloud existence that is intended to be given, but a two-way response (cloud or clear sky) is desired, then the probability values can be turned into a binary cloud distribution map by means of appropriate threshold values: if, in a given point, the probability of cloud presence exceeds a given threshold value, then it is considered to be a cloud, otherwise to be clear sky.

Difference from the pattern of polarization described by these models is, therefore, generally caused by some object, typically a cloud (or maybe an airplane or terrain feature). Polarization of skylight is basically caused by sunlight scattered once in the atmosphere. As a result of multiple scattering of sunlight in a cloud, cloud-light will be practically unpolarized (p « 0). The resultant polarization of a light beam propagating from a cloud is given by the superposition of unpolarized light of the cloud and polarized light typical of the singly scattered clear sky in the atmosphere between the cloud and the observer.

Light propagating from the direction of the cloud, therefore, comprises, on one hand, the unpolarized light (p « 0) of the cloud, as well as the partially linearly polarized light (0 < p < 1) scattered in the atmosphere between the cloud and the apparatus suitable for measuring polarization. At clear sky only the polarized light of the scattered skylight is visible (0 < p < 1). In cloudy and clear sky conditions, the same light scattering process will result partially linearly polarized (0 < p < 1) scattered light with identical characteristics. In case of clear sky, the apparatus for measuring polarization will only detect this polarized light, in case of cloudy sky the unpolarized light of the cloud will also mix in. Therefore, by knowing the degree of polarization of the light we should detect in a given cloudy point of the sky, if the given point of the sky was clear, we can calculate the intensity of light - having a degree of polarization belonging to the clear sky - should be subtracted from the actually observed light, in order to obtain totally unpolarized light, i.e. to have, after the subtraction, the unpolarized light of the cloud only. The intensity of light we need to subtract is proportional to the thickness of the atmosphere producing the scattered light, i.e. the distance of the cloud, as the polarized component of the measured light is as intensive as far the cloud is, because the more light scattering events may occur in the atmosphere between the cloud and the observer. Fig. 1 shows the intensity of unpolarized light in a given i color channel, which is l _ne utr,i = 2lmin,i, while the intensity of the polarized light is l _po i,i = Imaxj - Iminj = 2pjlj, where I, is the average intensity I, = (l _max ,i + l _m in,i)/2. If p, degree of polarization of the light is measured for a given i color at a given point, and we know that in that point the degree of polarization is p _c iear,i in case of clear sky, then from this we can determine an Xj relative distance, for which it is true that

Pi " Itotal.i ^— i ' Pciearj " lciearj ⁼ Pcloudj ' lcloudj ⁼ 0, which gives

Xi ⁼ Pi ^■ ItotaU /(Pclear.i · lciearj) _>

where l _to tai,i = Imax.i + Iminj and 0 < x, < 1 is the relative distance of the cloud measured in a given i color channel, x, = 0 means that the distance of the cloud is

0 m in the given color channel (it would belong to the case, wherein p, = 0, i.e. where the cloud is so close that there is not yet any scattered polarized light), x, =

1 belongs to the case, wherein the distance of the cloud is infinite, i.e. the sky is clear, pi · l _to tau = Pciearj · lciearj- In all three different color channels (red, green, blue) we have a \ degree of polarization and an l _to tau total intensity value. Total intensity, therefore, means the sum of the maximum and minimum intensities, which can be determined on the basis of the fitted function, and equal to the intensity which can be measured without the polarization filter. The p _c iearj degree of linear polarization of the cloudless, i.e. clear, sky and the l _c i _e ar intensity belonging to a cloudless sky (lciearj is the l _to t _a ij intensity value measured in the case of clear sky) are determined by means of the Rayleigh-model, Berry-model or the patterns of the degree and direction of polarization and the intensity values measured earlier from the clear sky. In addition to the detailed models, therefore, the earlier measured data for a given Sun-position can be used in the formula determining the relative distance of the cloud.

In certain embodiments of the method according to the invention, therefore, the degree of linear polarization of the cloudless sky and the intensity of the cloudless sky is determined by means of the Rayleigh-model, the Berry-model and/or the previously measured degree of polarization and values of intensity. Note that the Berry-model provides the patterns of the degree of polarization and the direction of polarization only, while the Rayleigh-model provides intensity distribution as well, in addition to the patterns of the degree of polarization and the direction of polarization. When using the Berry-model, for calculating the relative distance of a cloud, it is necessary to determine the intensity distribution of the clear sky from earlier measured values, by the Rayleigh-model or in any other way.

The application of the Berry-model is especially preferable in the course of the method according to the invention, as it provides a more accurate description of the polarization of the skylight than the Rayleigh-model, by also considering the unpolarized, neutral points of the sky, which are not considered by the Rayleigh- model. As a special case, the Berry-model reproduces the Rayleigh-model: in such cases when the angle distance between the neutral points and the Sun or anti-Sun is 0 degree.

In all three color channels, the respective distribution of the Xj relative cloud-base distance can be determined in the examined sky-portion (x _r in the red, x _g in the green, and x _b in the blue color channel). A total x relative cloud-base distance and its distribution in the sky can be received from their weighted average (x = a x _r + b-Xg + c-Xb). The values of a, b and c weights are determined on the basis of an optimization procedure by means of maximizing the accuracy of the final result. If the cloud-base distance measurement method according to the invention is performed on the zenith, i.e. above the observer, then the relative cloud-base distance obtained for the distance of the cloud is equal to the relative height of the cloud-base. In case of other directions, namely in non-vertical directions and when the axis of the recording device of the apparatus for measuring polarization is not vertical at the time of taking the image, then the relative distance of the cloud-base measured from the observer is obtained as relative cloud-base distance. Then, the relative cloud-base height corresponding to the given point can be calculated from a relative cloud-base distance received in the given direction by a geometrical calculation considering the curvature of the Earth's surface as well. In certain embodiments of the method according to the invention, therefore, a relative cloud- base height is calculated from at least a part of the relative cloud-base distances, preferably by means of geometry.

Note that the degree of polarization is affected by factors other than those mentioned above, such as the moisture and aerosol content of the atmosphere.

From the x relative cloud-base distance, we can deduce which cloud is more distant or nearer to the apparatus, therefore, a relative distance value can be calculated thereby. For determining the absolute distance of clouds, the relative distances received by means of the method according to the invention are to be calibrated at least in one point of the sky. Such calibration may be performed by the aforementioned LIDAR technology.

LIDAR can determine the absolute cloud-base distance in a single point. The LIDAR apparatus emits a high-intensity monochrome laser beam into the air, and then collects the laser light backscattered from the atmosphere with a telescope, and, from the intensity of this backscattered light and from the travel time, it is possible to deduce the atmospheric aerosol concentration and the base distance of any cloud if incidentally present. By means of a LIDAR apparatus, the algorithm for determining the relative cloud-base distance can be calibrated at each measurement, because, if the absolute cloud-base distance is known for a single point, then by using the relative cloud-base distances obtained by the method for measuring the cloud-base distance according to the invention, the absolute values of the distance of clouds can be calculated, within the limits of accuracy of the method, in each point recorded in the image. To conclude, in some embodiments of the method according to the invention, calibration is performed in such a manner, that an absolute cloud-base distance is measured for at least one point by means of LIDAR, and absolute cloud-base distance is determined by the absolute cloud-base distance taken by LIDAR for at least a part of the points of the images using the total relative cloud-base distances.

For calibration of the method according to the invention, i.e. for calculating the absolute values of cloud-base distance, a so-called stereo measurement can also be used instead of the LIDAR-based method. The calibration is, therefore, performed in certain embodiments of the invention in such a manner, that an absolute distance is measured for at least one point by means of stereo measurement, and using the absolute distance measured by the stereo measurement, absolute cloud-base distance is determined for at least a part of the points of the image using the total relative cloud-base distances.

For taking stereo measurements, two apparatuses suitable for measuring polarization or simple cameras (one of these will be referred to as image recording apparatus in the description of the stereo measurement) are installed at some 100 meter or 1-2 kilometer distance from each other, and the two image recording apparatuses are operated synchronously. The stereo measurement can be used as calibration for the method according to the invention, with the object to determine the absolute distance of the cloud located at least in one point of the sky, and, by calibrating the relative cloud-base distances determined by the method according to the invention with these absolute distance values, to determine in at least one part of the sky the absolute distance of points of clouds located therein.

Calibration by stereo measurement, i.e. determination of absolute distance is the solution of a relatively simple geometrical problem, illustrated in Fig. 2. Two images taken of the same object in space from two different optic angles are given. For a matter of simplicity, consider a single point P(x, y, z), the image of which appears in both pictures taken by the two apparatuses. The O origin of the coordinate system indicated in Fig. 2 is halfway in the x direction between the respective A and B objectives of the two image recording apparatuses, while the y axis is parallel with the imaging planes of the image recording apparatuses. Points A and B are at identical d distance from z axis in x direction, xi and x _r points are the positions of point P in the left and right images according to the local coordinate system, the origin of which is indicated as points O| and O _r, respectively. Both image recording apparatuses have identical f focal lengths. By observing the similar triangulars in Fig. 2, it can be seen that

xi/f = (d - x)/z and x _r /f = -(d + x)/z.

By substituting x:

(X, - x _r )/f = 2d/z.

The distance of point P is, therefore, in the two images: (xi - x _r ) = 2df/z. By using this expression, and knowing d and f values, we can determine the exact value of z based on (xi - x _r ), which means that the real z distance of point P measured from the image plane can be determined from the distances of point P measured in the two images. The z position of the image planes may differ (e.g. due to terrain), however since the difference is known, the calculation can be corrected.

Stereo measurement can only be performed for such points, which can be found in the respective images of both image recording apparatuses. This, on the one hand, excludes examination of points seen in the field-of-view of only one apparatus, as well as of those points that cannot be precisely separated or identified in both images (e.g. in the middle part of a homogenous gray spot). The most problematic part of stereo measurements is the identification of P points in the respective images of the two cameras. To facilitate this, pattern recognition, clustering as well as other algorithms are used, the applicability of which according to the invention are described herebelow.

Stereo measurements are, therefore, used for calibrating the relative cloud-base distances obtained by the method according to the invention. By means of determining the absolute distance of a single point by stereo measurement, the relative distance map can be used for determining the absolute distance of the other points. As a matter of fact, by means of calibration by stereo measurement, an absolute cloud-base distance map is obtained for the entire field of view from the relative cloud-base distances.

Calibration with LIDAR or by stereo measurement is, therefore, performed in a manner as follows. If the D _ab s(P) absolute distance and the D _re i(P) relative distance of a P point is known, then the D _re i(X) relative distance of any arbitratry X point can be transformed into an absolute value:

Dabs(X) = D _r e,(X) - D _abs (P) / D _re |(P).

A stereo measurement can be based on photometry only, i.e. based on the comparison of two different photos, nevertheless polarization images can also be used, especially preferably in cases, when the individual points or larger structures can be identified more easily therein.

In order to be able to reconstruct the spatial position of the objects, e.g. the individual points of the clouds, from the images captured from the two fields of view, first they must be identified in the images. If no background information were available about the picture, then this non-trivial, calculation-demanding procedure would have to be completed for the image information recorded by every single light detection element of the image recording apparatus. The calculation-demand can be exponentially decreased by decreasing the number of points to be calculated. To this end, a clustering algorithm making use of the results of preferably both polarimetric and photometric measurements may be used, which makes an estimation as to which light detection elements recorded the image information belonging to a continuous area (according to any one aspect), by way of example to a cloud. These areas may then be described by significantly less points, rendering the completion of the stereo measurement quicker by orders of magnitude (or making the calculation possible on simpler, slower computers as well). Any other algorithm determining the position of the clouds can be applied in this step, which is described together with the clustering algorithm herebelow. In certain embodiments of the method according to the invention, therefore, the stereo measurement is taken in such a manner that cloudy segments are determined in the linear polarization images, equivalent cloudy segments are identified in the respective images of the stereo measurement, and the absolute cloud-base distance is determined by means of the points equivalent with each other selected in one of the equivalent cloudy segments.

Determination of cloud position, e.g. by means of clustering algorithm, could preferably be part of the method according to the invention for measuring cloud- base distance. In one embodiment of the method according to the invention, therefore, cloudy segments are determined in the linear polarization images, and the points equivalent with each other are selected in the cloudy segments, i.e. the relative cloud-base distance is determined in the cloudy segments only. In certain embodiments of the method according to the invention the cloudy segments are defined as described hereabove by means of artificial intelligence. As artificial intelligence, evolutionary algorithm, neural networks or genetic algorithm is used. By doing so, significant calculation-demand may be spared consequently. At the same time, it should be noted, that the method according to the invention may also be completed for clear sky as described hereabove, in which case x«1 is obtained for relative cloud-base distance, which refers to the sky being clear. In the method according to the invention, the results of several independent measurements may preferably be used in order to improve accuracy. In light of the above, it is not necessary to determine in advance the position of the clouds as a separate step in the method according to the invention, but in the specially preferred embodiments of the invention, the data obtained by polarimetric and/or photometric measurements corresponding to the position of clouds can be used for determining the cloud-base distance.

In the stereo measurements according to the above, we are looking for matching | and x _r points. For this, pattern recognition algorithms are typically used, seeking such points in the image of the other camera, which show similar arrangement to that of the surroundings of the examined point. This can then be complemented with the extra information of algorithms of artificial intelligence (belonging to a cluster and to a cloud-type). The artificial intelligence algorithm used in a preferred embodiment of the invention, therefore, defines cloud cover by means of clustering. This algorithm can be preferably used according to the above as an initial step of the method according to the invention for measuring cloud-base distance so as to decrease the calculation-demand and/or during stereo measurements.

Unconventional methods of artificial intelligence are typically used in processing of measurement data. The methods can be very efficiently used in cases where input data and information is previously unknown, such as those typically produced in the often chaotic processes of nature.

The exemplary artificial intelligence-based algorithms to be used in the method according to the invention will be described herebelow. The evolutionary algorithm defines the clusters, i.e. the corresponding parts of the measured segment, on the basis of optical and spatial characteristics measured per light detection element. In this way, the areas with common characteristics can be managed together, thereby speeding up their respective processing, on the other hand enabling the individual continuous areas to be compared in a number of images recorded with spatial or temporal difference. This is a precondition for multi-camera distance measuring, such as the stereo measurement described hereabove.

The clustering algorithm applied in the method according to the invention considers the (geometric and photometric) distances of the individual points with different weights. A number of factors affect the appropriate weighting. In identifying unknown influencing factors, other algorithms of the field of artificial intelligence can also be used.

Neural networks imitate the structure of the nervous system. The measured data can be fed as input data to these networks, while outputs are generally categories according to given aspects, in which categories the input is then ranged. The structure of the network enables it to be trained for appropriate operation by an expert by means of known patterns. In the present case, this means setting appropriate parameters to the applied clustering algorithm.

For setting parameters of the clustering algorithm, the so-called genetic algorithm can be also used, which is also an evolutionary algorithm belonging to the field of artificial intelligence. For the operation of this (and many other algorithms belonging to the field of artificial intelligence), an expert has to define a function (fitness function), to determine how good is the solution of the problem. This, in many cases, is significantly simpler than finding the solution to the original problem. The genetic algorithm determines whether a solution to the problem is good, similarly to how genes work, by means of the fitness function. The algorithm starts with a number of arbitrarily generated gene versions. The genes may mutate (arbitrarily change), recombine (interchange gen-sections with one another) and proliferate according to the fitness function: the version giving a better solution to the problem may be copied, while those with worse solutions would have a greater chance of being removed. This way, after a time, those versions will get into majority, which solve the initial problem better. The ability of mutation of the genes ensures for the process to find a global optimum instead of a local optimum.

In the method according to the invention, neural networks and genetic algorithms can also be used in addition to the clustering algorithm for determining cloud types. This means classifying of already recognized cloud parts into cloud types, a task typically performed by an expert. The expert's viewpoints are, nevertheless, not entirely conscious, known or are rather difficult to algorithm. In this sub-task, the neural network trainable by patterns known to the expert or the genetic algorithm showing adaptive learning behavior with the help of the fitness function defined on the basis of the patterns will facilitate the appropriate use of latent parameters.

For determination of cloud position, in the method according to the invention, other, by way of example photometric, methods can be used, as well. The points of the image belonging to clouds can be determined by way of example as follows. If in a given point, |lb - l _r | < c l and |lb - l _g | < c lb, where l _r is the light intensity measured in the red spectral range, l _g in the green spectral range, and l _b in the blue spectral range, and c being a parameter, then a given point of the image will be grayish (achromatic), i.e. belonging to a cloud, while in the opposite case it would be a cloudless, i.e. clear (bluish) point. According to another, more simple photometric method, a given point of the image belongs to a cloud, if l _b /l _r < c\ where c' is again a parameter. Thus, this latter method does not consider green intensity.

Cloud position can be especially preferably determined according to the above on the basis of polarimetric information as well. In this we use the fact, that cloud light is almost unpolarized, the measured degree of polarization and the degree of polarization estimated by the theoretical models or the database relative to the clear sky will show differences. The maximum degree of polarization, at which a given point of the image is yet to be considered cloudy, is regarded as a threshold value, which depends upon the direction of the examined point relative to the Sun. This threshold value is constantly refined by an optimization (minimum search) method so as to enable the most accurate determination of cloud position, as there is a continuously expanding database of cloudy sky images available. In accordance with the above, we examine with the models, whether the rate of singly scattering due to the influence of clouds differs from the rate indicated by the model or the previous measurements relative to the clear sky.

The so-called minimum search method, by way of example, is as follows: in a number of color pictures of the sky variously covered with clouds, a number of persons are asked to recognize the clouds, which are selected on a computer screen. The individually recognized cloud points are then averaged, and the clouds presented thereby will be accepted as 'reality', which then are made to be recognized by means of various computer cloud recognition algorithms. The difference between 'reality' and the result recognized by a given algorithm is called an error (error may have another definition as well). By changing the parameters of the recognition algorithm, the error can be minimized. When the error is minimal, the parameters of the algorithm are regarded as optimum, and further these values are used for determining the position of clouds.

To sum up, the method according to the invention can use numerous algorithms for determining the position of clouds, which make use of various theoretical models or measured data. By way of example, there is a choice between the Rayleigh model and Berry model, however, the use of other models is also possible. Another example is the use of photometric and/or polarimetric information: in the case of photometric determination of cloud position, the color image of the sky (i.e. the light intensity pattern of the sky measured in the three spectral ranges) is exclusively used by the algorithm, searching the colorless points of the image, which may be clouds. In an exclusively polarimetric determination of cloud position, clouds are recognized based on the pattern of measured degree and direction of polarization of the sky, which patterns are in most cases taken in three color channels. In the case of the most general, and at the same time generally the most accurate combined determination of cloud position, cloud detection is performed by making use of both photometric as well as polarimetric information. In certain embodiments of the method according to the invention, therefore, cloudy segments are determined by comparison of the measured degree of linear polarization and/or measured intensity with the degree of linear polarization of the cloudless sky and the intensity of the cloudless sky, respectively. The recognition of cloud position may be improved by means of a so-called polarization-based dehazing procedure. The polarization-based dehazing is carried out on the basis of the aforementioned fact, that cloud light can be considered practically unpolarized, while it is mainly skylight scattered in the air that has polarized intensity. Hence, by removing the scattered component, visibility and contrast of unpolarized clouds can be improved, thereby making cloud position determination more precise. The polarization-based dehazing is completed as follows. Firstly, a PI polarized intensity is calculated for each light detection element, which is the product of p degree of linear polarization and I intensity: PI = l p. According to the most simple algorithm, then, in each spectral range (red, green, blue) PI polarized intensity is subtracted from I intensity per each light detection element, thereby we receive a pre-filtered, dehazed image, wherein cloud images have more contrast. As, however, the light scattered in the air is only partially linearly polarized, therefore, this simple method is not capable of removing all scattered components hindering cloud detection. The method may be further refined by defining a c factor for the polarized intensity of each individual light detection element in all three spectral ranges in such a manner that by subtracting c PI value from the I intensity of a given light detection element, the color image thus received would only contain the fully unpolarized component. The respective c factors of the individual light detection elements are determined on the basis of the pattern of the degree of polarization of the Berry-type clear sky model calculated on the basis of the actual position of the Sun at the time of the measurement, in view of the fact that the degree of polarization measurable at clear sky is equal to the degree of polarization of light scattered in the air.

According to the above, therefore, by making use of the measured polarization information, i.e. degree of polarization and intensity, of skylight, the contrast of the color image of cloudy sky can significantly be improved in such a manner that light intensity in a given spectral range is decreased in a rate proportional to the degree of polarization. Namely, the contrast is decreased by the partially linearly polarized light scattered in the atmosphere between the cloud and the observer (e.g. a device for recording images) mixing with the light from the cloud. This disturbing polarized component can be partially removed from the skylight by means of the above algorithm, thereby receiving a color image of the cloudy sky as if the air between the observer and the clouds did not contain vapor and aerosols. Clouds can be detected with increased accuracy by means of any of the conventional photometric algorithms in the color image, the contrast of which we have thus increased. The extra polarization information, therefore, can be preferably used for indicating cloud presence and for determining cloud position, thereby improving the accuracy of cloud detection.

In preferred embodiments of the method according to the invention, the circularly polarized component of incident light is also measured. From the circular polarization images taken by means of a circular polarizing filter, further information relative to the clouds can be obtained. Based on the circular polarization images, we can distinguish the water clouds and the ice clouds. The degree of circular polarization determinable from the circular polarization image of the light from clouds consisting of water droplets is insignificantly small, while that of clouds consisting of ice crystals is higher. By setting an appropriate threshold value of the measured degree of circular polarization it can be decided whether the observed cloud is a water cloud or an ice cloud.

With this procedure we can also distinguish water and ice clouds from dust clouds. As the degree of circular polarization of light reflected from dust grains may exceed that of light scattered from ice grains, therefore, a second appropriate threshold value of the measured degree of circular polarization can be set so as to determine whether it is a dust cloud or water/ice cloud. Thus, in certain embodiments of the method according to the invention, the method according to the invention is completed by also taking a circular polarization image of the sky- portion by means of a circular polarizing filter, and determining in some points of the circular polarization image the degree of circular polarization of the incident light beam arriving in the point, and deciding based on the degree of circular polarization whether there is a water cloud, ice cloud or dust cloud in the point.

From the degree of circular polarization of skylight, it is possible to deduce the aerosol concentration thereof. The higher the degree of circular polarization, the higher the aerosol content of the atmosphere is. As the determination of the cloud- base distance is based on the degree of linear polarization of skylight, therefore, the measurement of cloud-base height is barely affected by aerosols.

The method for measuring the cloud-base distance according to the invention may be carried out in the images taken by way of example with an apparatus 1 demonstrated in Fig. 3 or with apparatus 1' demonstrated in Fig. 7.

The apparatus V demonstrated in Fig. 7 belongs to the prior art, however, the method according to the invention can be carried out by means of this apparatus 1 ' as well. The apparatus V takes three linear polarization images with a recording device fitted with three linear polarizing filters of different transmission directions as well as wide-angle optical devices 10', 10", 10"'.

As the sky is appropriately far, the relative cloud-base distance can be obtained, with high accuracy, by means of the apparatus 1 ', i.e. from the images captured by the recording device fitted with three different wide-angle optical devices 10', 10", 10"', nevertheless, the method according to the invention is especially preferably based on linear polarization images synchronously taken by a recording device, by way of example a camera, using at least three linear polarization filters with different transmission directions. If the images serving as the basis for the method according to the invention are not taken synchronously, then the clouds might move in the time elapsing between taking of the images, thereby distorting the results of the method for measuring cloud-base distance.

The method for measuring the cloud-base distance according to the invention can be carried out in such a manner that linear polarization images are captured through at least three linear polarizing filters with different transmission directions from essentially one position and at essentially the same time. Such apparatuses shall henceforward be referred to as apparatus for measuring polarization.

The apparatus 1 for measuring polarization according to the invention is illustrated in Fig. 3. The apparatus 1 takes the required polarization images with a, by way of example digital, camera 12 fitted with wide-angle optical means 10. In the apparatus 1 , polarizing filters (not illustrated in Fig. 3) are arranged in the light path of the camera 12 between the wide-angle fisheye lens being a part of the wide- angle optical device 10 and the CCD-detector (not illustrated). In the apparatus according to the invention, three polarizing filters with different transmission directions (such as 0°, 45°, 90° or 0°, 60°, 120°) can be placed in the light path. The three linear polarizing filters, by way of example, can be complemented by a circular polarizing filter, thereby ensuring measurability of ellipticity, i.e. the degree of circular polarization, the use of which is described hereabove, in addition to light intensity and the degree and direction of linear polarization. If the spectral sensitivity of the image recording device capturing the image, by way of example a CCD-detector, i.e. if the respective sensitivity curve extends into the infrared region, then infrared measurements can also be taken.

In the apparatus 1 according to the invention, a sun-shielding disk 16 can be moved around the center of the wide-lens objective 10 in horizontal (azimuth) and vertical (elevation) direction with an adjustment means 20 and an adjustment means 8, respectively. In order for the sun-shielding disk 16 to cover the least possible field-of-view, it is to be placed as far as possible from the wide-angle optical device 10, while it is important for the wide-angle optical device 10 to be fully shielded by means of the sun-shielding disk 16. The devices comprised by the apparatus 1 are arranged in an instrument box 22.

The sun-shielding disk 16, therefore, is to be moved in the direction of the Sun so as to keep the wide-angle optical device 10 fully shielded for the sake of eliminating the image corrupting affects of disturbing inner light reflections. As one option, a picture of the sky may be taken without the use of the sun-shielding disk 16 prior to the polarization measurement. In this image, the position of the Sun is recognized by searching for the maximum light intensity, and prior to the actual measurement, the sun-shielding disk 16 is moved to this position. Another option is determining the Sun's position in the sky by means of the astronomical library freely available on the internet in view of the geographical coordinates, the direction of the apparatus 1 relative to the geographical north and the exact time of the measurement, and the sun-shielding disk 16 is moved to this position in advance. To determine the exact geographical position and date of measurement, a GPS (Global Positioning System) may be arranged in the instrument box 22. If, therefore, the geographical position of the polarization measurement is determined by GPS, then the position of the Sun and the desired position of the sun-shielding disk 16 can be determined by the above method. If necessary, data of a built-in magnetic compass can also be read, if, for example, in case of a ship-based instrument, the direction of north cannot be made out of the GPS data due to sideways sea currents.

The apparatus 1 comprises a transparent weatherproof cap 14, arranged on the upper portion of the instrument box 22 as seen in Fig. 3. The wide-angle-lens optical device 10 looks at the sky through the large enough transparent weatherproof cap 14 made of plastic or glass. Preferably, the sun-shielding disk 16 together with the adjusting means 18 and adjusting means 20 are arranged inside of the weatherproof cap 14. As much as possible, the area around the wide-angle optical device 10 within the weatherproof cap 14 is homogenous matte black for the sake of decreasing the disturbing reflections. Cooling of the instrument box 22 of the apparatus is provided by Peltier-elements, arranged with large enough heatsinks between the inside of the box and the outside world. It is extraordinary important that the instrument box 22 has low internal humidity in order to ensure moisture-free inner surface for the transparent weatherproof cap 14. For this purpose appropriate amount of silica gel can be applied.

In the apparatus according to the invention, the pictures are captured synchronously, moreover, according to the invention the error arising from different fields-of-view of the images is also minimized, and in certain embodiments it does not occur. Herebelow some embodiments of the apparatus according to the invention are presented, which differ in the position of the polarizing filters.

In the apparatus 1 according to the invention, the three images are taken by means of a single wide-angle optical device 10 and a camera 12. In one embodiment of the apparatus according to the invention, a beam splitter prism system 30 (not illustrated in Fig. 3) is placed into the path of an incident light beam 34 arriving through the wide-angle optical device 10. Fig. 4 illustrates the beam splitter prism system 30, the path of the incident light beam 34, and an array of polarizing filters 29 comprising linear polarizing filters 26, 26', a further polarizing filter 26" shown only in Fig. 5, and a circular polarizing filter 28, arranged in front of an image recording device 24. Fig. 4 shows the path of the incident light beam 34 and its components, indicated by thick dashed line. Continuous line indicates the visible edges of the beam splitter prism system, thin dashed line indicates the covered edges. It can be seen, that the incident light beam 34 is split into four segments by means of light splitting planes, such as beam splitting plane 31 , thereby the visual information held by the incident light beam 34 enters the image recording device 24 through the linear polarizing filters 26, 26', 26" and the circular polarizing filter 28 of the array of polarizing filters 29. In order to the components of the incident light beam 34 have identical length of optical path in a medium of different refractivity, i.e. in the beam splitter prism system 30, namely to have identical visual information on the linear polarizing filters 26, 26', 26" and the circular polarizing filter 28, in the path of the components of the incident light beam 34 there are arranged optical-path-length compensating means. Such optical-path- length compensating means are, by way of example, optical-path-length compensating means 32 in Fig. 4 as well as path-length compensating means 32' in Fig. 5. In the arrangement illustrated in Figs. 4 and 5, the optical path-length to the image recording device 24 is identical for each component of the incident light beam 34. The present arrangement can ensure polarization images to be taken at the same time from exactly the same position, namely without any temporal or spatial difference between the images.

In certain embodiments of the apparatus according to the invention, therefore, a beam splitter prism system 30 is arranged in the path of the incident light beam 34 transmitted through the wide-angle optical device 10, wherein the polarizing filters 26, 26', 26" are arranged to cover neighboring partial areas of the image recording device 24. The polarizing filters 26, 26', 26", 28 preferably have a form similar to that of the image recording device 24. Furthermore, the beam splitter prism system 30 splits the incident light beam 34 into a number of incident light beam components corresponding to the number of partial areas covered by the different polarizing filters 26, 26', 26", 28, and the beam splitter prism system 30 transmits the incident light beam components to the partial areas.

Fig. 5 illustrates the beam splitter prism system 30 in a view different from that of Fig. 4. Fig. 5 shows the linear polarizing filters 26, 26', 26" and the circular polarizing filter 28 comprised by the array of polarizing filters 29 in greater detail. By way of example, the linear polarizing filters 26, 26', 26" have transmission axes of 0°, 45°, 90° or 0°, 60°, 120° respectively, while the circular polarizing filter 28 is right handed or left handed. By the use of the present embodiment of the apparatus according to the invention, one embodiment of the method according to the invention is carried out in such a manner that the image is taken by the camera 12, and the incident light beam 34 passing into the camera 12 is being transmitted by means of the beam splitter prism system 30 to the areas fitted with different polarizing filters 26, 26', 26", 28 arranged on the image recording device 24 of the camera 2. In the embodiments shown in Figs. 4 and 5, the surface of the image recording device 24 is split into four non-overlapping adjacent areas, the information recorded in the areas covered by the four different polarizing filters is evaluated independently, i.e. the effective resolution of the image recording device 24 is one-fourth of the resolution of the original image recording device in terms of the number of pixels. Conventional macroscopic polarizing filters 26, 26', 26", 28 may be placed in the light-paths. According to the above, the recorded images have the same field of view, and while being taken, do not require synchronization. The beam splitter prism system is to be made to suit the size of the image recording device 24 and of the imaging lens of the wide-angle optical device 10. The image recording device 24 may be a color-sensitive image recording device, such as one made color-sensitive by a Bayer-filter, or a color-insensitive image recording device. As mentioned hereabove, cloud-base distance may be determined on the basis of visual information comprising color (e.g. containing three color channels) or based on visual information not comprising any color. Fig. 6 illustrates the positioning of Bayer-filter, where the Bayer filter 38 is arranged directly overlapping the detection surface of the image recording device 36 comprised of light detection elements. In Fig. 6, by way of example, the Bayer-filter is arranged with three different color filters. For example, one piece of blue color filter 39, two pieces of green color filter 39' and one piece of red 39" color filter make up a unit of four. The pattern of the Bayer-filter 38 is illustrated by differently hatched lines per color filters 39, 39', 39".

The method according to the invention is preferably performed by an apparatus 1 according to the invention for measuring polarization, which comprises an optical device 10 producing a light beam 34 of a sky-portion, an image recording device 24, 36 capable of receiving the light beam 34 and taking an image thereof, at least three linear polarizing filters 26, 26', 26", 42, 42', 42" with different transmission directions, and the polarizing filters 26, 26', 26", 42, 42', 42" are arranged affecting different parts of the light beam 34, which parts of the light beam are projected onto a common image recording device 24, 36.

Note that in certain embodiments of the apparatus according to the invention, the image recording device is a digital image recording device 24, 36 with light detection elements arranged on its detection surface. However, the apparatus according to the invention can be preferably implemented with analogous color- sensitive or color-insensitive image recording devices as well, and the method according to the invention can be suitably applied to analogous images as well. The method according to the invention is preferably executed by means of an apparatus according to the invention comprising a circular polarizing filter 28, 44 arranged overlapping at least one part of the detector surface of the image recording device 24, 36. In certain embodiments of the apparatus according to the invention, therefore, there is an array of color filters, preferably a Bayer-filter 38, arranged between the polarizing filters 26, 26', 26", 42, 42', 42", 44 and the image recording device 24, 36, overlapped manner covering the detector surface of the digital image recording device 24, 36. The perspective view of a further preferred embodiment of the invention is also identical with that of apparatus 1 , i.e. the three images are recorded by means of a single wide-angle optical device 10 and a camera 12. In this embodiment, an array of polarizing filters 40 is arranged in the path of the incident light beam. The array of polarizing filters 40 is illustrated in Fig. 6, however, the incident light beam is not shown. In this embodiment, the image is recorded by image recording device 36, which has been made color-sensitive in the present embodiment by means of a Bayer-filter. Another embodiment is also possible, which is different from the present embodiment only in that it does not contain a Bayer-filter 38, therefore, the arrangement is not color-sensitive. The method according to the invention is carried out in the aforementioned embodiments of the apparatus according to the invention in such a manner that the image is taken by the camera 12 having an array of polarizing filters 29, 40 placed in front of its image recording device 24, 36.

In the present embodiment, the image holds polarization information because of the use of a polarizing filter 40. Such an array of polarizing filters 40 is used comprising three linear polarizing filters 42, 42', 42" with different transmission directions and a circular polarizing filter 44, thereby forming units of fours in the polarizing filter array 40. The polarizing filters 42, 42', 42", by way of example, have transmission directions of 0°, 45°, 90° or 0°, 60°, 120° respectively, and the circular polarizing filter is either right handed or left handed. An embodiment of the array of polarizing filters is also possible, wherein another combination of polarizing filters is used, by way of example, a combination of linear polarizing filters only. It is evident from the figure, that one color-filter 39, 39', 39" of the Bayer-filter, arranged as blue, red or green color filter in the present embodiment, belongs to one light detection element 37 of the image recording device 36. In this way, four light detection elements of the image recording device 36 are comprised in an effective chromatic light detection element. In a way as shown in Fig. 6, one linear polarizing filter 42, 42', 42" or one circular polarizing filter 44 belongs to one such effective chromatic light detection element, i.e. to the four light detection elements of the image recording device 36 arranged in the form of a square. In the present embodiment, the resolution of the image containing the polarization information is one-fourth of the image containing color information, i.e. one- sixteenth of the resolution of the image recording device 36, in terms of the number of pixels. The polarized image in this embodiment is made up of individual light detection elements covered by the linear polarizing filters 42, 42', 42" of identical transmission directions and the circular filter 44. In this embodiment, it will evidently not be realized in an exact manner, that the images recorded with the individual polarizing filters are taken from the same one position, i.e. with identical fields-of-view; the adjacent light detection elements recording different polarization information, therefore, do not look at the same point of object. In areas of the object-space with high intensity gradient, this could lead to a small degree of distortion in the results. The arrangement used in the present embodiment, however, minimizes the field-of-view difference and thereby indirectly the error as well. In this embodiment, the use of a more complex polarizing filter array 40 is required, as opposed to the aforementioned embodiment, wherein the polarizing filter array 29 was used comprising of four related polarizing filters 26, 26', 26", 28. On the other hand, in the polarizing filter array 40, there is a number of polarizing filter areas corresponding to the light detection elements of the image recording device 36. The advantage of the present embodiment of the invention is that it does not require the use of the beam splitter prism system 30. In accordance with the above, in some preferred embodiments of the apparatus according to the invention, the polarizing filters 26, 26', 26", 28, 42, 42', 42", 44 are formed as an array of polarizing filters 29, 40. In the preferred embodiments of the apparatus according to the invention, at least a part of the components under the weatherproof cap of the apparatus 1 is painted matte black in order to avoid reflections disturbing the imaging of the incident light beam, similarly to known apparatuses.

In the embodiments using the Bayer-filter, the resolution may be refined by means of a software. There are numerous algorithms for software-based refinement. Many times these are incorporated into the software of the camera, for example, the color image of a camera need not be separately processed per color, but a ready-made color picture can be downloaded from the camera. The most simple software refinement is based on interpolation, where, by way of example, in a red light detection element of the Bayer-filter, green intensity is given by the interpolation of the intensity of the adjacent green light detection elements. These algorithms, therefore, process the image of the Bayer-filter based color cameras, and the resolution of the final color picture comes near to the physical resolution of the image recording device. Fig. 7 shows a known polarization cloud detection apparatus V suitable for carrying out the method according to the invention. In the apparatus 1 ', there are three or four cameras 12', 12", with image recording devices and wide-angle optical devices 10', 10", 10"', the camera of the wide-angle optical device 10"' is not shown, arranged according to the given application. In the method according to the invention, images are captured by the apparatus V by means of cameras 12', 12" positioned having essentially the same fields-of-view, and having polarizing filters placed in front of the cameras. The required three polarization images of the sky or the four polarization images of the sky including the circular polarization image as well are captured at the same time, simultaneously with the apparatus 1 '. The three or four cameras 12', 12" are preferably placed as close to each other as possible, e.g. in case of three cameras 12', 12" they are placed in a triangular form, while in the case of four cameras, they are placed in a square form, however other arrangement might also be possible. The optical axes of the wide-angle optical devices 10', 10", 10"' of the cameras 12', 12" are as parallel as possible. In order to avoid any possible differences, the images of the cameras 12', 12" and of those other camera(s) not shown are to be calibrated, and the measurement results are to be corrected accordingly. The image recording devices are to be synchronized so as to ensure that the desired polarization images are taken at the same time.

Sun-shielding disk 16' of the apparatus V is similar to the sun-shielding disk 16 of apparatus 1 , nonetheless, because of the three or four adjacent cameras 12', 12", the sun-shielding disk 16' is of a greater size, as it has to keep all three wide-angle optical devices 10', 10", 10"' or four wide-angle objectives in full shadow at the same time. The position of the sun-shielding disk 16' may be adjusted with the adjusting means 18' and adjusting means 20'. The apparatus 1 comprises a weatherproof cap 14' and the devices of the apparatus are being arranged in instrument box 22'.

The invention is, of course, not limited to the preferred embodiments described in details above, but further variants, modifications and developments are possible within the scope of protection determined by the claims.