The invention relates to a system and method for detection and localization of moving objects at medium and long distances.

Collisions of objects such as animals and drones with wind turbines, aircraft and other high value installations can lead to accidents which are endanger wildlife and also cause damage to infrastructure. However, collisions can often be prevented if the moving objects are detected early enough. There is therefore both an environmental and industrial benefit to the reliable detection and accurate localization of moving objects.

Stereovision provides a technique to monitor and locate moving objects in the ground and the air. However, prior art systems are limited in that they have a fixed field of view in which the optical axes of the cameras are perpendicular to the base line between the cameras. This creates a fixed observation area and objects approaching from different angles cannot be detected unless a further pair of stereoscopic cameras is installed or a specific rotating system moving whole large baseline is installed.

It is therefore an object of the invention to make the range of observation area defined as a stereovision field of view more flexible and decrease the devices’ size, without compromising configurability of a pair of stereoscopic cameras.

According to the invention there is provided a system for determining the distance, Dk of a detected object to a line through the centres of two cameras. The system comprises a first camera C1 , a second camera C2, a line between centres of the first and second camera forming a baseline B, and a processing system configured to receive data from the first camera C1 and the second camera C2 to determine the position of an object, when the first and second cameras are oriented with their optical axes with an angle a to the base line.

Contrary to conventional systems this allows a much greater degree of flexibility in the position of the cameras as the optical axes of the cameras need not be perpendicular to the base line between the cameras’ centres. The cameras may be positioned adjacent to each other horizontally, vertically above each other or even at an angle (no-vertical and non-horizontal) to each other. However, the optical axes of the cameras need to be parallel. Furthermore, the cameras may be rotatable such that they can detect and localize a moving object coming from any direction and therefore a greater field of view is possible.

The first and second cameras have an angular field of view, (p ₀ and a resolution, y ₀ along Y1 and Y2 axes. The first and second cameras have first and and second axes respectively which are perpendicular to the base line B and the optical axis of the respective camera. The Y1 and Y2 axes initially lie on the same line as the base line B and are within the image plane of the respective camera and perpendicular to the first and second axes respectively. The pixel’s number of the centre of the object projection on the image plane of the first camera C1 along a Y1 axis is yi and the pixel number of the centre of the object projection on the image plane of the second camera C2 along a Y2 axis is y ₂ . The processing system may be configured to use angles cp ₀ , yo, the length of baseline, B and angle a, and pixel numbers yi and y ₂ to determine the position of an object.

According to some embodiments the distance Dk is given by:

The first and second cameras and the processing system may be connected to a network and such an arrangement allows remote processing of the data.

The first camera C1 may be rotatable about its axis X1 defined as an axis perpendicular to Y1 and lying within the centre of the image plane of the camera and the second camera C2 may be rotatable about an axis X2 parallel to the axis X1 and lying within the centre of the image plane of the camera C2. This enables objects to be viewed around the X axes.

The first camera C1 may be rotatable about an axis Y1 which is perpendicular to the axis X1 and lies on the base line B and the second camera C2 may be rotatable about an axis Y2 which is parallel to the axis Y1. This enable objects to be viewed around Y axes.

If the cameras rotate there may be a control system configured to control the cameras to rotate the first and second cameras equally such that the object remains in the stereovision field of view.

The processing system may be configured to determine the position of the object in three dimensions based on the determined distance, D and the position of the object on the image plane in at least one of the first camera C1 and the second camera C2. An accurate three-dimensional position is helpful to take evasive action should it be necessary. If a three-dimensional position of a moving object is determined at more than one point in time then its velocity can be determined which can aid evasive action still further.

According to the invention there is provided a system allowing bird mortality mitigation as described above, wherein the first camera C1 and the second camera C2 are arranged substantially vertically above each other.

According to the invention there is provided a method for determining the distance, D, of an object from a first camera C1 , the method comprising receiving first image data from a first camera C1 , receiving second image data from a second camera C2, determining the distance of an object from a first camera C1 , the distance being calculated by: wherein (p ₀ is the angular field of view of each of the cameras, y ₀ is the resolution of each of the cameras along the Y1 and Y2 axes, yi is the number of the pixel which represents the object’s centre projection on the image plane of the first camera C1 on the Y1 axis, y ₂ is the number of the pixel which represents the object’s centre projection on the image plane of the second camera C2 on the Y2 axis, B is a length of base line between the first and second cameras’ optical axes and a is the angle of the cameras relative to the base line. The first camera is the camera which is closest to the object.

The position of the object in 3D space may be calculated based on the determined distance, D and the projection of the object’s centre on the image plane at least one of the first camera C1 or the second camera C2.

According to the invention there is provided computer program code means configured to carry out the steps as described above.

Figures 1a and 1 b depict a basic configuration of a pair of stereoscopic cameras according to the invention;

Figure 2 illustrates an angular transformation of the stereovision system coordinates when optical axes of the cameras are not perpendicular to the base line; and

Figure 3 depicts a system architecture and its components according to the invention.

The invention is a customized stereovision monitoring system for locating distant objects. The system along with the algorithmic methods enables an extraction of the 3D features from images. In classic stereovision, the cameras’ optical axes are perpendicular to a baseline (where the baseline is the line segment connecting cameras’ centres) and then cameras’ image planes are placed on the same Euclidean plane. The invented configuration is characterized in that the cameras’ optical axes are set at an angle (a) with respect to the base line (which can be horizontal or vertical or neither horizontal nor vertical) in such a way that the image planes of the invented system are placed on two parallel planes. The range of angle 0°< a <360°.

Figure 1 b depicts a camera basic set up according to the invention. There is a first camera C1 which has an image plane 11 and a second camera C2 which has an image plane I2. The cameras are arranged along a baseline B between the two camera centres and the optical axes of the cameras are set at angle a wherein a is substantially non-perpendicular. Directions of X1 and X2 are arranged perpendicular to the base line between the cameras and within the image planes of the cameras.

To extract the 3D features from images, the stereovision coordinate system is transformed using its geometric features. The transformation is carried out in relation to the first camera C1 in such that the coordinate system and the scene are shifted by the cameras rotation angle a, as shown in Figure 1 and Figure 2. The mathematical model needed for computing the object’s distance from the vision system is presented below.

The distance of the detected object from the line on which the base line lies can be computed using the following formulae:

The second part of the equation can be simplified as:

Then, D can be defined as:

This enables the distance of objects to be calculated when the cameras’ optical axes are not arranged perpendicularly to the base line B. This means that cameras do not need to be arranged along lines perpendicular to the direction of detection and so a greater range of configurations of cameras is possible.

Furthermore, in an embodiment both the first camera C1 and the second camera C2 are arranged to rotate around the X1 and X2 axes. The first camera C1 and the second camera C2 are configured to be rotated in so that their optical axes are both at the same angle a to the base line.

The description above refers to the cameras being rotated around X1 and Y2 axes and the cameras’ optical axes being at a non-right angle a to the baseline. However, they could instead, or additionally, be rotated around the Y1 and Y2 axes (namely the other axes of the image planes of the cameras) and similar calculations could be carried out. The resulting distance calculation would be:

The cameras are depicted as being arranged side by side, but they could equally be arranged vertically. Such an arrangement would be particularly suitable near a tower of wind turbine as the cameras could be located on the wind turbine itself. The present invention may be used to detect objects in the vicinity of airfields or wind turbines or even very tall buildings. It may be used to detect any flying object but particularly birds or drones. The detection of such objects allows evasive action to be taken if necessary.

Although the invention is described with respect to cameras detecting visible light in the visible range, infra-red cameras could also be used to detect moving objects at night.

Figure 3 depicts the basic hardware configuration of the monitoring system according to the invention. The system comprises two image capturing devices. As can be seen the first image capture device comprises a camera C1 and a local processing unit 10. The second image capture device comprises a camera C2 and a local processing unit 20. Both image capture devices are connected, via a network 30 to a processing system 40. The network 30 may be hardwired network or a wireless network. Images captured by the cameras can therefore be processed locally on a local processing unit or remotely on processing system. Within the invention the cameras are coupled in a stereoscopic mode, although the cameras may operate independently and separately. Each camera can be equipped with an additional local processing (CPU or/and GPU) unit applying an embedded object detection algorithm.

Preservation of moving object detected at medium and long distances and its 3D localization require real time processing of high-resolution images. To reduce the computational costs, the monitoring system may be implemented as a distributed computing system. In this option, each of image capturing devices would work independently as a standalone subsystem sending acquired pre-processed information to the processing system 40. Based on data from two image capturing devices, the processing system computes the 3D coordinates of the detected objects.

The image capturing devices can work in a synchronous or asynchronous mode. When they work in the synchronous mode, the processing system triggers image capturing so that images are captured by both cameras at the same time. If the image capturing devices work in the asynchronous mode as standalone units, the processing system adjusts data of images based on the timestamp specified by network time protocol.

Various further aspects and embodiments of the present invention will be apparent to those skilled in the art in view of the present disclosure.

“and/or” where used herein is to be taken as specific disclosure of each of the two specified features or components with or without the other. For example, “A and/or B” is to be taken as specific disclosure of each of (i) A, (ii) B and (iii) A and B, just as if each is set out individually herein.

Unless context dictates otherwise, the descriptions and definitions of the features set out above are not limited to any particular aspect or embodiment of the invention and apply equally to all aspects and embodiments which are described.

It will further be appreciated by those skilled in the art that although the invention has been described by way of example with reference to several embodiments. It is not limited to the disclosed embodiments and that alternative embodiments could be constructed without departing from the scope of the invention as defined in the appended claims.