STEWART, Scott (631 Drake Center, Winnipeg, Manitoba R3T 5V4, CA)
PALIWAL, Jitendra (631 Drake Center, Winnipeg, Manitoba R3T 5V4, CA)
PRYSTUPA, David, A. (631 Drake center, Winnipeg, Manitoba R3T 5V4, CA)
STEWART, Scott (631 Drake Center, Winnipeg, Manitoba R3T 5V4, CA)
PALIWAL, Jitendra (631 Drake Center, Winnipeg, Manitoba R3T 5V4, CA)
1. A method of separating defective kernels from a quantity of grain comprising: providing a quantity of grain kernels, at least some of said kernels suspected of defects; isolating respective individual kernels from said quantity of grain kernels; subjecting said respective individual kernels to visible light or near infra-red analysis and determining visible light or near infra-red scattering from said respective individual kernels; and rejecting said respective individual kernels if the scattering is above a threshold level.
2. The method according to claim 1 wherein the kernels are wheat.
3. The method according to claim 1 wherein the defect is Fusarium head blight infection. 4. The method according to claim 1 wherein the threshold level is derived from visible light or near infra-red scattering from a non-defective kernel.
5. The method according to claim 1 wherein the threshold level is derived from visible light or near infra-red scattering from a set of non-defective kernel.
FUSARIUM DETECTION METHOD PRIOR APPLICATION INFORMATION
The instant application claims the benefit of US Provisional Patent Application 60/911 ,311 , filed April 12, 2007.
BACKGROUND OF THE INVENTION
Fusarium head blight (FHB) is a fungal disease that affects small cereal grains which may cause the grain to become shrivelled, light in weight and infused with the myoctoxin deoxynivalenol (DON). This toxin can cause sickness in humans and accordingly there are limits on the parts-per-million (ppm) DON allowable in end products made from wheat based on their intended use. For example, below 1 ppm is acceptable for human consumption, below 5 ppm for swine and below 10 ppm for cattle or chickens.
Delwiche and Gaines (2005, Applied Engineering in Agriculture 21 : 681-688) teach the use of single- and two-wavelength linear discriminant models using visible and near-infrared wavelengths for developing monochromatic and bichromatic high-speed grain sorters.
PCT Application WO03/025858 teaches an algorithm for interpreting color images of seeds which comprises filtering interferences, extracting features and transferring the features to a trained neural network for classification.
US Patent 5,898,792 describes a method for determining flour yield, protein content and bulk density of cereal kernels by using a trained neural network.
US Patent 5,956,413 teaches a method to orient kernels and the application of neural networks to classify kernels from image analysis.
US Patent 6,427,128 teaches an image analysis method to detect defects in granular objects using reflected and transmitted light.
US Patent 6,845,326 teaches a specific setup of a spectrophotometer in transmission mode to measure the constituents in grain.
SUMMARY OF THE INVENTION
According to a first aspect of the invention, there is provided a method of separating defective kernels from a quantity of grain comprising:
providing a quantity of grain kernels, at least some of said kernels suspected of defects; isolating respective individual kernels from said quantity of grain kernels; subjecting said respective individual kernels to visible light or near infra-red analysis and determining visible light or near infra-red scattering from said respective individual kernels; and rejecting said respective individual kernels if the scattering is above a threshold level.
BRIEF DESCRIPTION OF THE DRAWINGS
Figure 1 is a schematic diagram of one embodiment of the invention. Figure 2 is a schematic diagram of another embodiment of the invention.
Figure 3 is a schematic diagram of another embodiment of the invention.
Figure 4 is a schematic diagram of an ejection apparatus of the invention.
Figure 5 is a schematic diagram of a separation apparatus of the invention. Figure 6 is a graph of signal versus time.
Figure 7 is a plot showing the rejection of infected kernels.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, the preferred methods and materials are now described. All publications mentioned hereunder are incorporated herein by reference. Described herein is a method of detection, singulation and separation of damaged wheat kernels.
While in a preferred embodiment, the kernel is damaged by an FHB infection, it is important to note that using the method described herein, any fungal
infection or imperfection that changes the bi-directional reflectance distribution function of the sample can be detected. For example, a fungal infection on, or near, the surface of a grain kernel that affects a substantial fraction of the surface area (>50%) compared to a healthy kernel will in general satisfy this condition. It is of note that other imperfections both natural and environmental that satisfy this condition, that is, damage a significant or substantial or major portion of the surface area of the kernel, for example, frost damage, can be detected using the described method.
According to an aspect of the invention, there is provided a method of separating defective kernels from healthy kernels in a quantity of grain comprising: providing a quantity of grain kernels, at least some of said kernels suspected of defects; isolating respective individual kernels from said quantity of grain kernels; subjecting said respective individual kernels to visible light or near infra-red analysis and determining visible light or near infra-red scattering from said respective individual kernels; and rejecting said respective individual kernels if the scattering is above a threshold level. The threshold level may be the scattering for a single healthy kernel or may be an average from a set of healthy kernels. In other embodiments, there may be greater scattering from a damaged kernel compared to a healthy kernel, meaning that when scattering for a respective kernel is above a certain level or is a certain percentage above a known control, the kernel is rejected as damaged. The percentage may be 40-60% more scattering. According to an aspect of the invention, there is provided a method of separating a defective kernel from healthy kernels in a quantity of grain comprising: providing a quantity of grain kernels, at least some of said kernels suspected of defects; isolating an individual kernel from said quantity of grain kernels; subjecting said individual kernel to visible light or near infra-red analysis and determining visible light or near infra-red scattering from said individual kernel; and rejecting said individual kernel if the scattering is above a threshold level.
As discussed below, the kernels may be wheat although as will be appreciated by one of skill in the art, kernels from other suitable grains may be used or analyzed or sorted as well.
As discussed below, the threshold level may be derived from infra- red scattering from a non-defective kernel or from a set of non-defective kernel.
As will be appreciated by one of skill in the art, the method involves the steps of separating or isolating individual kernels of grain and then forwarding the respective individual kernels for visible light or near infra-red analysis. As discussed below, the inventors have discovered that defective kernels, for example, fungus-infected kernels, scatter visible light or near IR radiation differently than healthy kernels. Accordingly, as discussed below, this property may be used to separate defective kernels from healthy kernels. Also provided are examples of optical arrangements for visible light or near IR analysis as well as apparatuses for singulating or separating kernels for visible light or near IR analysis and apparatuses for automated rejection of defective kernels. It is however important to note that these examples are provided for illustrative purposes and a wide variety of separation, detection and rejection arrangements may be used under the invention.
In a preferred embodiment, a kernel of grain suspected of infection or substantial imperfection, for example, surface imperfections is isolated and passed through a tube having an entry end, an exit end and a port. The port is arranged to accept a light source.
The light source may be for example an LED, a laser, or an incandescent, halogen, fluorescent or flash lamp. In some embodiments, a LED in combination with collimating optics is the preferred light source due to low power, long life and spectral characteristics. A laser works very well, but has a shorter operational life and is more easily damaged but is suitable in some embodiments and for some applications. Other light sources including incandescent, halogen, fluorescent and flash lamps will work in combination with collimating optics, but the lamp life and energy costs are less favourable. Wavelengths between 400 nm and 700 nm were tested and healthy kernels can be distinguished from infected kernels over the whole range except between 475 and 525 nm. Accordingly, wavelengths between 400-475 nm and between 525-700 nm may be used in the invention.
In most embodiments, a collimating optic is placed after the light source to minimize the angular divergence of the output beam. In some embodiments the angular divergence can be 30 degrees, but in preferred embodiments the angular divergence is less than 5 degrees and most preferred less than 1 degree. It is noted that the design of such collimating optics is well within ordinary skill in the art. Alternatively, a commercially available collimating optic, for example, such as those available from for example Thorlabs, or Edmund Optics may be utilized within the invention.
In all embodiments, the output beam is incident upon a kernel to be inspected. In one embodiment, the diameter of the incident beam at the kernel is slightly larger than the kernel so that the entire surface of the kernel can be examined at once. For example, for wheat kernels, a beam having a diameter between about 3 mm and about 5 mm, preferably about 4 mm will generally meet this requirement. In these embodiments, it is preferred that the intensity distribution of the incident beam should be as uniform as posssible. The difference between the intensity at the centre of the beam and the intensity at the edge of the beam should be less than 50% and preferentially less than 10%. In other embodiments, the incident beam is substantially smaller than the kernel and is scanned across the surface so that an infection affecting only a small part of the surface can be detected. In these embodiments, a laser with a Gaussian beam profile may be used because the time average illumination at each point will be the same regardless of beam profile.
In one preferred embodiment (Figure 1), the output beam is produced by a source S, collimated by a collimating optics C and then passes through a port in the side of a tube and is reflected by a small prism (P) to follow the central axis of the tube. The radius of the output face of the small prism, R2, preferentially matches the radius of the collimated output beam so that the distance H2 can be minimized. In another embodiment, a small mirror also is positioned to reflect the collimated output beam along the central axis of the tube at the position P. Light from the collimated beam is incident on the kernel to be tested and light is scattered by the kernel into a plurality of angles ranging from 0 to 90 degrees. A lens, L, is situated behind the prism or mirror which collects and focuses scattered light on a detector D1 that is immediately preceded by a band
pass filter Fl The maximum acceptance angle for the collection tube, A2, is arctan(R1/H1) and the minimum acceptance angle, A1 , is arctan(R2/(H1 +H2)). H1 , H2, R1 and R2 are chosen such that the minimum acceptance angle, A1 , is approximately 30 degrees and the maximum acceptance angle, A2, is approximately 60 degrees. H1 is preferably in the range of 1 mm to 5 mm, R1 is preferably in the range of 5 mm to 13 mm. The inner walls of the tube are reflective. The tube may have a constant diameter, or in a preferred embodiment the tube is tapered such that the diameter is smaller at the sample end than at the detector end of the tube as shown in Figure 2. The angle A3 is preferably in the range from 15 to 30 degrees. A key feature of this arrangement is that light scattered at small angles will impinge upon the prism or mirror and thus be prevented from travelling further along the axis of the tube. As noted above, R2 defines the minimum scattering angle transmitted by the tube. The diameter and vertical displacement of the tube proximate to the kernel defines the maximum scattering angle transmitted by the tube.
Figure 3 shows another embodiment in which light is generated at source S and optionally collimated at C1. The light is transmitted by a tube, optical fibre, or a plurality of optical fibres, E, and is optionally collimated at C2 prior to incidence upon the kernel to be tested, K. An optically opaque ring of radius R2 surrounds the inner conduit proximate to the kernel. The scattered radiation is collected by a plurality of optical fibres G between radii R2 and R1 that transmit the scattered radiation to detector D1 , immediately preceded by a band pass filter (not shown). The inner, R2 and outer, R1 radii for the collection fibres G are chosen so that the minimum acceptance angle is approximately 30 degrees and the maximum acceptance angle is approximately 60 degrees. In some embodiments light collected by the optical fibres G may be transmitted to a single detector. The intensity measured by the detector is compared with a previously determined threshold value to determine whether the kernel is good or bad. The threshold value is determined by comparing signals from a training set of kernels classified as healthy or infected. In other embodiments the light collected by optical fibres G are mapped onto a plurality of detectors and the pattern of intensity values is compared with previously determined patterns to determine whether the kernel is healthy or infected. The patterns for comparison are delermined by comparing
signals from a training set of kernels classified as healthy or infected.
In another embodiment of Figures 1 , 2 and 3, the detector D1 is a Raman spectrometer. In this embodiment, the elastically-scattered Rayleigh component contains information about the surface texture and the inelasitically- scattered Raman component contains information about the chemical composition at the surface of the kernel. In this embodiment, the Raman spectrum is compared with a training set that includes possible contaminants such as urea, pesticides, fertilizer, insect parts, fungicides, herbicides, rat poison, and the like. It is notable that the preferred wavelength, 470 nm, is close to the Argon laser line at 488 nm that is widely used by those skilled in the art for exciting Raman spectra.
We have also used a mirror with a small hole in the centre to collect scattered light. The mirror is set at an angle of approximately 45 degrees. All that is required is that the reflected light is not collinear with the incident beam, so any angle in the range of 20 to 70 degrees would also be practical. Incident light passes through the small hole in the mirror, drilled at the angle of incidence to preserve the beam profile, and is scattered by the kernel. Scattered light reflected from the mirror is collected by a lens system and analysed. Experiments showed that the signal from the mirror however were not as good as the results from a direct backscatter measurement. As will be appreciated by one of skill in the art, at least the portion of the tube proximal to the kernel must be substantially round and the inner walls must be polished. The minimum diameter of the tube is set by the height of the tube above a kernel, the desired solid angle for measurement, and the size of the required input optics (approx 3 mm). The tube acts as a waveguide beyond the first bounce, so the walls must be polished to minimize attenuation. The main constraint on the upper size of the tube (or fibre bundle) is the spacing between kernels. The upper diameter of the tube is preferably 50 mm or less and most preferably 25 mm or less.
As will be appreciated by one of skill in the art, the light source must produce a collimated beam. If the beam is not collimated, the angular distribution of the scattered radiation is convoluted with the angular distribution of the incident radiation. The measurement window is about 30 degrees, so a reasonable tolerance is 2 degrees or 35 mrad.
In a preferred embodiment, the scattered light is focussed by a lens at the top of the tube onto a detector. Alternately, the lens focuses scattered light into an optical fibre that transmits the light to a detector, or to a specific location on a detector array. Alternately, if a fibre bundle is used in the place of the tube (Figure 3), the fibre bundle terminates at a detector.
Additionally, experiments have shown that the width of the scattered radiation peak does not correlate well to kernel size. Infected kernels are smaller on average yet produce broader peaks. The important diagnostic parameter is the scattering into the angular range given above per unit of surface area. Healthy kernels have a larger surface area, but scatter less intensely. The product of area * intensity per unit area for healthy and infected kernels is similar. The contrast is 0-15% without normalization and 40-50% with normalization. Although the area of a kernel can be measured with an imaging system, the same result can be obtained with less cost and complexity by measuring the shadow cast by a kernel as it passes between a light source and a detector. In a preferred embodiment, the light source S emits two distinct wavelength bands, which travel collinearly through the optical system as shown in Figures 1-3. As one skilled in the art will realize, light from two sources may be combined with a beamsplitter, a prism, or a grating to make a collinear beam. Light scattered at a first wavelength is collected as shown in Figures 1-3, with the optional addition of a band pass filter F1 to block the second wavelength. Light of the second wavelength is either blocked by the kernel K or transmitted around the edges and detected by a second detector D2 after passing through band pass filter F2. If the light source is uniform, then the fraction of light blocked will be proportional to the area of the kernel. If the light source is not uniform, a line scan detector proximate to and immediately underneath the kernel may be used to detect the positions at which the transmitted light level is below a pre-determined threshold value. The width of the kernel is proportional to the number of dark pixels and the area is calculated by adding successive rows. The second wavelength is chosen to have minimal attenuation by the substrate B supporting the kernel and a large attenuation by the kernel. The substrate supporting the kernel B 1 may be chosen to act as a band pass filter. In some embodiments, the two wavelengths may be the same. In a preferred embodiment, the second wavelength is longer than the first wavelength to reduce the effect of
scattering by dust and is chosen for transmission through the substrate material B Wavelengths between 600 and 1000 nm are suggested
As can be seen in Figures 6 and 7, healthy and infected kernels have different distributions separated by more that the sum of their standard deviations Experiments with a wide variety of wheat samples have demonstrated that Fusarium infected kernels scatter 40-60% more radiation than healthy kernels per unit area Generally, Fusarium infected kernels tend to be smaller than healthy kernels
As will be appreciated by one skilled in the art, a variety of means may be used to deliver individual kernels to the analysis tube As will be apparent to one skilled in the art, multiple tubes may be used in parallel, each arranged to accept kernels from the separator In these embodiments, the system comprises a plurality of tubes, a separator and sorting means wherein kernels are either accepted or rejected based on scattering properties, as discussed herein For example, in some embodiments, the separator comprises a wheel having a plurality of slots, each slot being sized and/or arranged to allow a single kernel to pass through the respective slots In some embodiments, the slots may be tapered to allow the kernels to fall into the slots
In some embodiments, a hopper H, feeds kernels through a vibrating grate G as shown in Figure 5 A rotating cylinder C has n rows of m slots, where n and m are integers There are n such slots arranged around the circumference of the cylinder Four are shown in Figure 5, but the number can be in the range of for example 1 to 20 and for example preferably 4 to 8 The radius of the cylinder is in the range of 6 to 25 mm and preferably about 12 mm Each slot is just large enough to accept one kernel For wheat, the slots are preferably 3-4 mm long, 2-3 mm wide and 2-3 mm deep As the cylinder rotates, a kernel is transported from an input port to an output port Upon exiting the output port, the kernel falls onto a belt B travelling in a direction parallel to the axis of the cylinder C Each time the slots line up with the output gate, a row of m kernels is transferred to the moving belt In some embodiments, an inclined slide may be placed between the output port and the belt to accelerate kernels to a velocity close to the velocity of the belt In a preferred embodiment, the kernels are transferred from the output port to a slot in the moving belt The velocity of the belt and the rotation of the cylinder are
synchronized so that kernels are transferred to the belt at regular intervals. For wheat, the interval between slots on the belt is between 3 mm and 100 mm and preferably between 6 mm and 8 mm.
In some embodiments, kernels are transferred to m consecutive positions on the belt at once. In other embodiments the transfer from cylinder to belt is interleaved. In some embodiments, the m slots along the cylinder are in phase. In other embodiments the slots along the cylinder are out of phase by an integral divisor of 2*pi.
In a preferred embodiment, there are a plurality of parallel cylinders feeding a plurality of tracks in a belt. In one embodiment, there are 128 cylinders feeding 128 parallel tracks on a belt. In this example, the cylinder with n=8 and m=100 rotates at 30 revolutions per minute and transfers 400 kernels per second to the belt, that has a slot spacing of 6 mm and speed of 2.4 m/s. In this case the belt will transport approximately 5530 kg of wheat per hour. In some embodiments, grain may fall through a spout onto a spinning conical section. As the grain falls along the conical section, the number density decreases such that single kernels fall into tapered slots at the perimeter of the conical section.
In another embodiment, grain is fed into a system consisting of a plurality of belts moving at successively higher speeds. Grain may be fed onto a first belt moving at speed vl The first belt accelerates the grain to speed v1 before transferring the grain to a second belt moving at speed v2. Assuming the belts have equal width, the mass density on the second belt will be m2 = m1*v1/v2, where ml is the mass density on the first belt and m2 is the mass density on the second belt. The process may be repeated until the mass density corresponds to a fractional monolayer of kernels. The last belt may have a pattern of tapered slots into which individual kernels may fall.
In some embodiments, the tapered slots on the belt receiving individual grain kernels may have a small hole with a diameter less than the diameter of the smallest grain kernel received by the system. After receiving a grain kernel, the slot passes a detector as described above. The detector may be caused to produce a signal signifying that the kernels is either accepted or rejected. If the kernel is rejected, the kernel may be removed from the belt by one
of the following mechanisms:
1. The belt or wheel may have slots with grain kernels on one side and holes that line up with the slots on a reverse side. In a preferred embodiment, a tube mates with the hole in the wheel or belt assuring alignment. Upon receiving a signal from the detector to eject a kernel, a small rod may be driven through the hole to displace a grain kernel. In a preferred embodiment, the rod may be driven by a solenoid.
2. A quantity of gas sufficient to eject a kernel may by pass through the hole in the bottom of the slot by activating a valve connected to a reservoir of compressed gas.
3. Heat energy sufficient to vaporize a volatile liquid within a small chamber is produced by a resistive element as shown in Figure 4. A detector D senses the proximity of a slot containing a kernel. This signal is logically combined in an AND gate with the boolean result from the defect detection apparatus described previously at logic unit L. In a preferred embodiment the logic unit is a CLPD. In another embodiment the logic element is a FPGA. In another embodiment, the logic element is a micro-controller. If the kernel is to be rejected and the slot is aligned, an electric current flows through resistor R vaporizing a volatile liquid. The gas produced travels into tube T displacing the rejected kernel K into a hopper H. The volatile liquid is replenished from a reservoir by a duct W. In a preferred embodiment, the liquid is water and an electric current supplied to a resistive element produces the heat. For wheat a kernel the quantity of water is approximately 6 E-4 g and the energy supplied is approximately 1.6 J. 4. In yet another embodiment, an electric charge is applied to infected kernels and healthy kernels are grounded to eliminate any net charge. After a charge is applied the kernels are made to leave the belt and travel along a parabolic trajectory toward two bins. An electric field is applied in a direction parallel or anti-parallel to the belt to accelerate charged kernels to a first bin whilst healthy kernels fall into a second bin. In a related embodiment, the charge may be applied to healthy kernels and infected kernels are grounded. In prior work, whole and broken wheat kernels were sorted on the basis of a difference in their charge to mass ratio (2004,
Abdel-Salam and El-Kishky. Electrical Insulation and Dielectric Phenomena, 2004. CEIDP '04. 2004 Annual Report Conference on, p 377-380) All of the kernels were charged. However, in this embodiment, kernels are selectively charged or uncharged.
While the preferred embodiments of the invention have been described above, it will be recognized and understood that various modifications may be made therein, and the appended claims are intended to cover all such modifications which may fall within the spirit and scope of the invention.