Improved methods for determining the sex of chicks.

Abstract

The invention relates to a method for determining the sex of a chick, comprising: determining (100) a region of interest in the image in which the feathers of a wing are visible, and running, on said region of interest, a classification model (400) trained on a training data set comprising images of male chick wings and female chick wings, in order to determine whether the chick is male or female.

Description

METHOD FOR IMPROVED DETERMINATION OF THE SEX OF A CHICK Technical field

[0001] The present invention relates to a method for determining the sex of chicks by observing the arrangement of the feathers at the end of the wings, and a device for implementing this method. Prior art

[0002] It is known to determine the sex of chicks based on the arrangement of the feathers, located at the tips of the wings. In fact, as can be observed in Figures 1a and 1b, the tips of the wings include two rows of feathers which are the primary remiges and the primary covert feathers. In the figures, the ends of the diets and covert feathers, respectively, have been highlighted. In male chicks, in Figure 1a, these two rows of feathers end at the same level, while in female chicks, in Figure 1b, the end of the primary remiges ends further than the end of the primary coverts, and therefore an alternation of relatively short and long feathers can be observed.

[0003] The determination of the sex of the chicks by feathers was first carried out manually by an operator, which requires a significant amount of time.

[0004] Solutions have also been proposed to automate this determination.

[0005] Thus, document EP 1 092 347 discloses a method for determining the sex of chicks which uses a conveyor to separate the chicks and bring them to the level of a camera. At this point, the chick is unbalanced, for example by vibrations of the conveyor, to cause it to spread its wings of its own initiative, and one or more images of the chick's wings are acquired.

[0006] Image processing methods have also been proposed to automatically determine the sex of the chick from the acquired image. For example, in US 6,396,938, a first method comprises extracting shape parameters of each feather (length, middle position and tip position), and calculating a parameter that quantifies the variation in length of two adjacent feathers.

[0007] Another method described in this paper involves locating the coordinates of the feather tips, and determining a polynomial function connecting the feather tips. The sex of the chick is determined based on the parameters of the polynomial function.

[0008] Chicks for which the result is uncertain can be studied manually or re-determined automatically.

[0009] However, in document US 6,396,938, these methods are implemented on a previously annotated image of a chick's wing to identify at least the tips of the feathers. The methods described therefore do not allow for the implementation of a completely automatic processing for determining the sex of the chick from the capture of the image. Summary

[0010] In view of the above, the invention aims to provide an improved and fully automatic solution for determining the sex of a chick from an image of a chick's wing.

[0011] In particular, the invention aims to propose rapid and reliable processing of an image to determine the sex of the chick.

[0012] Another object of the invention is to provide a fully automated process for determining the sex of the chick.

[0013] In this regard, the subject of the invention is a method for determining the sex of a chick, the method being implemented by computer from an image of a chick, the method comprising: determining a region of interest of the image on which the feathers of a wing are visible, implementing, on said region of interest, a classification model trained on a training data set comprising images of male chick wings and female chick wings, to determine the male or female sex of the chick.

[0014] In embodiments, the method is implemented for each of a plurality of images acquired on the same chick, and further comprises a step of determining the sex of the chick from the results obtained by the classification model for all of the images.

[0015] In embodiments, determining a region of interest of the image comprises: scanning the image with a window of determined size to define a plurality of regions of the image, for each region, calculating a Haar characteristic of the region, applying a trained classifier to each Haar feature to determine whether or not the region represents feathers, and determining a region of interest in the image as a region representing feathers.

[0016] In embodiments, the method further comprises processing the region of interest to determine a set of lines corresponding to the feathers of the chick in the image, determining a set of parameters from the extracted lines, and the classification model is applied to said set of parameters.

[0017] In embodiments, processing the region of interest to determine a set of lines corresponding to the feathers in the image comprises: performing edge detection processing on the region of interest, and applying a Hough transform to the edges resulting from the processing to determine a set of lines corresponding to the feathers visible in the region of interest.

[0018] In embodiments, the contour detection processing comprises: implementing a Gaussian filter and thresholding the region of interest to obtain a binary representation of the region of interest, calculating a distance map on the binary representation of the region of interest, to determine a distance between each point and a contour closest to said point, and normalizing said map to obtain a grayscale representation of the region of interest, and implementing an erosion operation on the obtained grayscale representation.

[0019] In embodiments, thresholding the region of interest comprises determining, for each current pixel of the region of interest, a thresholding value determined based on the intensity values ​​of the pixels included in a local neighborhood of the current pixel.

[0020] In embodiments, determining the set of lines representing the feathers further comprises: rotating the region of interest by an angle determined from the angle, relative to the horizontal, of the longest lines, to make said lines substantially horizontal, and the deletion of lines which, after rotation of the region of interest, extend in a direction forming an angle greater than a predetermined threshold with respect to the abscissa axis.

[0021] In embodiments, determining the parameters from the extracted lines includes: identifying a set of lines corresponding to long feathers, and identifying a set of lines corresponding to short feathers.

[0022] In embodiments, the method comprises performing a rotation of the region of interest so that the lines representing the feathers extend substantially horizontally, ranking each line in order of length, and identifying the set of lines corresponding to long feathers comprises: initializing the set of lines corresponding to long feathers, said set comprising the longest line, performing, for each line included in said set, the following steps: o identifying the set of lines neighboring the line considered along the vertical axis, o calculating, for each neighboring line, a difference in length and a distance between the center of the neighboring line and the center of the line considered, o if the relative difference and the distance are less than respective thresholds, identifying the neighboring line as a line corresponding to a long feather,and adding to the set of lines corresponding to long feathers.,

[0023] In embodiments, the identification of a set of lines corresponding to short feathers comprises implementing, for each line corresponding to a long feather of the set, starting with the line located at the maximum vertical position of the set, the following steps: identifying, among the lines not belonging to the set of lines corresponding to long feathers, the lines neighboring the line considered, calculating, for each neighboring line, a difference in length, a distance along the vertical axis between the line considered and the neighboring line, and a distance along the horizontal axis between a distal end, respectively proximal of the line considered and the proximal end, respectively distal of the neighboring line, if the calculated differences and distances are less than respective thresholds, the identification of the neighboring line as a line corresponding to a short feather.

[0024] In embodiments, the parameters determined from the lines include at least: a number of lines corresponding to long feathers, a number of lines corresponding to short feathers, an average angle between the lines and the horizontal, and an average gap, measured vertically, between two adjacent lines.

[0025] In embodiments, the parameters determined from the lines further comprise at least one of the group consisting of: minimum, maximum and / or average horizontal and / or vertical position of the centers of the lines, minimum, average and / or maximum distance, horizontally and / or vertically, between the center of two consecutive lines, minimum, average and / or maximum length of the lines corresponding to long feathers, minimum, average and / or maximum length of the lines corresponding to short feathers, average pixel intensity of a line corresponding to a short feather, average pixel intensity of a line corresponding to a long feather, difference in average pixel intensity between the lines corresponding to long feathers and the lines corresponding to short feathers, location relative to the horizontal axis of each line, a minimum, average and / or maximum angle between two lines, a minimum distance,average and / or maximum between the proximal ends of two lines corresponding to a long feather and a line corresponding to a short feather located between them, a minimum, average and / or maximum distance between the distal ends of two lines corresponding to a long feather and one line corresponding to a short feather located between them.

[0026] In embodiments, the classification model trained to determine the sex of a chick is a decision tree.

[0027] In embodiments, the classification model is trained on a database of annotated training images, where each training image is obtained by applying the steps of determining a region of interest and processing the region of interest to extract a set of lines representing the feathers, and the annotation includes an indication of the sex of the chick and an associated certainty level, determined from a number of lines corresponding to long feathers and a number of lines corresponding to short feathers.

[0028] In embodiments, the method is implemented on a set of images of the same chick, and includes determining the sex of the chick from the most frequently provided result by the classification model.

[0029] According to another object, a computer program product is described, comprising code instructions for implementing the method according to the preceding description, when executed by a computer.

[0030] The invention also relates to a device for determining the sex of a chick comprising at least: a camera adapted to acquire at least one image of a chick, and a computer configured to implement the method according to the preceding description on the image acquired by the camera.

[0031] In embodiments, the camera is adapted to acquire images in a wavelength range between 340 and 500 nm, preferably between 400 and 450 nm.

[0032] In embodiments, the device further comprises a conveyor adapted to bring a chick into the field of view of the camera, wherein the conveyor is adapted to unbalance the chicks such that the chick has its wings spread when in front of the camera.

[0033] In embodiments, the camera is configured to acquire a series of at least 20 images of each chick.

[0034] In embodiments, the device comprises a conveyor, a first station for detecting chicks of a first sex, male or female, comprising said camera, and an actuator adapted to pick up or eject from the conveyor the chicks detected as belonging to the first sex, in which the computer is configured to implement on the image acquired by the camera a first classification model optimized to detect the first sex, and the computer is further configured to implement a second classification model optimized to detect the second sex, on images acquired on chicks not having been determined to be of the first sex.

[0035] The proposed method makes it possible, from one or more images of a chick, to automatically, quickly and reliably determine the sex of the chick. The method notably comprises determining a region of interest in the image on which the feathers of the chick's wing are visible, then processing the image to automatically extract a set of lines corresponding to the feathers, and finally classifying the sex of the chick according to parameters of these lines.

[0036] The sex of a chick can thus be determined in 400 ms, which is much faster than previous methods. Brief description of the drawings

[0037] Other features, details and advantages will become apparent upon reading the detailed description below, and upon analyzing the attached drawings, in which: Fig. 1a

[0038] [Fig. 1a] already described, represents an example of a male chick wing. Fig. 1b

[0039] [Fig. 1b], also already described, represents an example of a female chick wing. Fig. 2a

[0040] [Fig. 2a] schematically represents an example of a device for determining the sex of a chick according to one embodiment. Fig. 2b

[0041] [Fig. 2b] schematically represents another example of a device for determining the sex of a chick according to one embodiment. Fig. 3

[0042] [Fig. 3] schematically represents the main steps of a method for determining the sex of a chick according to one embodiment. Fig. 4a, Fig. 4b

[0043] [Fig. 4a], [Fig. 4b] represent examples of images used for training a classifier used for determining the region of interest of the image. Fig. 5

[0044] [Fig. 5] shows an example of a Haar feature that can be calculated on the image for the determination of the region of interest. Fig. 6a

[0045] [Fig. 6a] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 6b

[0046] [Fig. 6b] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 6c

[0047] [Fig. 6c] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 6d

[0048] [Fig. 6d] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 6e

[0049] [Fig. 6e] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 6f

[0050] [Fig. 61] represents a step of an example of processing implemented on a region of interest of an image to extract a set of lines corresponding to the feathers. Fig. 7

[0051] [Fig. 7] represents a set of lines corresponding to long and short feathers of chicks. Description of the embodiments

[0052] With reference to Figure 3, a method for determining the sex of a chick will now be described. This method is implemented by a computer 10 which comprises one or more computers 11 of the type, for example, processor, microprocessor, controller, microcontroller, FPGA, etc., and a memory 12 storing code instructions executed by the computer. As will be seen in more detail below, the memory 12 can also store pre-trained classification models which are used during processing.

[0053] The method for determining the sex of a chick is implemented by processing one or more images representing a wing of the chick received by the computer 10.

[0054] In an embodiment shown schematically in Figure 2a, a device 1 for determining the sex of a chick comprises a camera 20 adapted to acquire images of chicks and a computer to which the camera 20 transmits the acquired images.

[0055] The camera 20 may be adapted to acquire images in wavelengths between 340 and 500 nm, for example, which is a wavelength range for which the best contrast is obtained for viewing wing feathers. The camera may be adapted to acquire images in a wavelength range between 400 and 450 nm. The camera may be monochromatic and centered on a wavelength in this range, for example 405 nm.

[0056] In embodiments, the device 1 comprises a conveyor 30 adapted to bring chicks individually and successively into the field of vision of the camera 20. The conveyor 30 may be adapted to unbalance the chick so that the chick has its wings spread when it passes in front of the camera. By way of non-limiting example, the conveyor 30 may comprise a portion located upstream of the camera 20, this portion being able to vibrate so as to unbalance the chicks.

[0057] In addition, the camera can take a series of images of each chick, for example at least 20 images acquired consecutively, for example 40 images of each chick, which the camera 20 transmits to the computer 10 for implementing the method for determining the sex of the chick. All of the images taken by the camera have the same dimensions and the same resolution.

[0058] In an embodiment shown in Figure 2a, the device 1 may comprise a single station for sorting the chicks, which comprises a camera 20 taking images of the chicks and the computer 10 directly determines whether the chick corresponds to a male or a female, by applying the processing described below which notably comprises the implementation of a classification algorithm trained to determine whether the chick is a male or a female.

[0059] Alternatively, such as for example the case shown in Figure 2b, the device 1 may comprise a first station comprising a camera 20 taking images of the chicks (with the wings deployed as described above), and transmitting these images to a computer 10 making it possible to determine only whether the chick corresponds to a first sex, male or female. If necessary, the computer 10 implements a classification model optimized to detect the first sex. In the example shown in Figure 2b, the first station makes it possible to determine only female chicks.

[0060] The first station further comprises an adapter adapted to collect, or eject, from the conveyor the chicks identified as corresponding to the first sex.

[0061] For the remaining chicks, in one embodiment, the computer 10 can apply to the corresponding images a second classification model optimized to detect the second sex. Alternatively, the device comprises a second station also comprising a camera for taking images of the remaining chicks. The computer 10 (which may be separate or identical to that of the first station) implements the processing described below on the images obtained and applies a model optimized to detect the second sex. The device 1 can comprise a second actuator adapted to collect or eject from the conveyor the chicks identified as corresponding to the second sex.

[0062] The remaining chicks can be collected for return to the starting point of the conveyor or for analysis by an operator.

[0063] According to yet another variant, the conveyor comprises two image-taking stations, each with at least one camera, making it possible to double the chances of good presentation of the chick with open wings, and a classification model between male and female sexes, or two classification models optimized respectively to detect each sex, receive as input the images from the two cameras.

[0064] Returning to Figure 3, the method comprises a step 100 of determining, from an image of the chick, a region of interest of the image on which the feathers of a wing are visible. This step makes it possible on the one hand to discard the images on which the feathers of the wings are not visible, and on the other hand to make it possible to implement the rest of the processing on a region which may be smaller than the original image, which therefore makes it possible to reduce the calculation time.

[0065] From the region of interest of the image thus obtained, the method further comprises, and as described in more detail below, the implementation of a classification model 400 trained on a training database comprising images of male chick wings and female chick wings, to determine the male or female sex of the chick. As indicated above, the method may comprise the successive implementation of a first model optimized to determine that a chick belongs to a first sex, male or female, and a second model optimized to determine that a chick belongs to the other sex, female or male.

[0066] In embodiments, the determination 100 of the area of ​​interest can be implemented by applying a trained model, for example by a deep learning type model.

[0067] In a particular embodiment, determining the region of interest of the image comprises scanning the image with a window of determined size, to define a plurality of regions of the image.

[0068] For each region obtained, Haar characteristics are calculated 110, making it possible to obtain a vector associated with each region. With reference to Figure 5, a Haar characteristic is obtained from a pattern M of pixels of a size smaller than the size of the region R. The pixel pattern comprises two subsets of pixels corresponding to determined locations within the pattern. The Haar characteristic is calculated by subtracting the cumulative intensities of the pixels of the first subset from the cumulative intensities of the pixels of the second subset. Thus, a Haar characteristic is obtained for each possible position of the pixel pattern within the region, and the vector obtained brings together all the Haar characteristics calculated over the entire region considered.

[0069] The obtained vector is then provided to a trained classifier 120 to determine whether the region from which the vector was obtained is a region representing feathers or not. With reference to Figures 4a and 4b, the classifier is advantageously trained by an Adaboost type algorithm, on an image database including images representing feathers (Figure 4a) and other images (Figure 4b) not representing feathers, on which Haar feature vectors have been calculated according to the previous paragraph. The classifier trained in this way can, for example, be a decision tree.

[0070] The region of interest of the image is the region detected as representing feathers by the classifier.

[0071] Multiple regions of interest may be detected in the same image. In embodiments, the following steps of the method may be implemented for all regions of interest detected for an image. Alternatively, the following steps of the method are implemented for a first region of interest, and only in the case where a determination of the sex of the chick is not possible from this first region of interest, the method is repeated on a second region of interest, and so on until a determination of the sex of the chick is obtained from the image.

[0072] In embodiments, once the region(s) of interest have been identified for an image, the method may comprise the direct application 400 of a classification model trained on said region of interest, for example of the neural network type.

[0073] Alternatively, and as shown in Figure 3, the method comprises, before implementing the classification model, a step 200 of processing the region of interest to determine lines corresponding to the feathers of the chick, and a step 300 of determining a set of relevant parameters provided as input to the model.

[0074] With reference to figures 6a to 6f, we will describe a mode of implementing a processing 200 of the (or each) region of interest to extract lines corresponding to the feathers of the chick on the image.

[0075] This processing includes the implementation of a contour detection 210 on the region of interest, and the determination of a set of lines 220, corresponding to the feathers, from the detected contours.

[0076] Figure 6a shows an example of a region of interest of an image obtained at the end of step 100 described above. In embodiments, the contour detection processing comprises the implementation of a Gaussian filter and thresholding of the region of interest to obtain a binary representation of said region, as shown in Figure 6b, i.e., a region in which all pixels are either black or white.

[0077] The thresholding is advantageously adaptive so as to compensate for possible variations in the lighting conditions of the image. The local threshold can be determined, for each current pixel of the region of interest, from the intensity values ​​of the pixels included in a local neighborhood of the current pixel, for example a square window centered on the current pixel. The threshold can for example be the average of the intensities of the pixels in the window.

[0078] The edge detection processing can then include a step of calculating a distance map on the obtained binary representation, to determine a distance between each point of the binary region and an edge closest to the point. The metric used to calculate the distances can for example be the Techbychev distance or chessboard distance. The obtained distance map is then normalized to obtain a grayscale representation of the region of interest, like the example shown in Figure 6c.

[0079] The processing then includes an erosion operation of the region of interest obtained, making it possible to reduce the noise in the image and in particular between the feathers.

[0080] The determination 220 of the lines corresponding to the feathers can then be carried out on the region of interest obtained by applying a Hough transform. In the example shown in Figures 6a to 6f, the lines obtained are visible in Figure 6d. It is understood from the preceding description that the lines are sets of pixels arranged substantially according to a line shape, but the lines are not necessarily perfectly straight or of constant width.

[0081] Once the lines are obtained, the method may also include rotating 230 the region of interest so that the lines are substantially horizontal. In this regard, the rotation angle may be determined by identifying the longest feathers in the region of interest, for example the two or three longest feathers, and calculating the angle of each line relative to the abscissa X-axis, and calculating the average angle over the feathers considered.

[0082] In embodiments, once the rotation has been performed, the angle of each line corresponding to a feather relative to the X axis can be recalculated, and the lines forming an angle, relative to this axis, greater than a determined threshold can be deleted, because these lines then correspond to noise and not to real feathers. The angular threshold can be between 20 and 30°, for example equal to 25°. An example of the result is shown in Figure 6e.

[0083] Returning to Figure 3, the determination of the parameters 300 from the lines representing the feathers firstly comprises the identification 310 of the lines corresponding to long feathers, and the identification of the lines 320 corresponding to short feathers. In Figure 6f, the LL lines corresponding to the long feathers are represented, and the LC lines corresponding to the short feathers.

[0084] The identification of the lines corresponding to long feathers 310 is implemented by initializing a set of the lines corresponding to long feathers, said set comprising the longest line of all those which appear on the region obtained in step 200.

[0085] The set is then completed with other lines which can be determined as follows: for each line included in the set of lines corresponding to long feathers, the identification of the set of lines neighboring the line considered, according to the vertical axis or Y axis. In this regard, an example is shown in Figure 7 in which the Y axis is represented and the lines are indexed from top to bottom, the indexes being represented to the right of each line. The number to the left of each line represents the ranking of the lines in order of length. So in the example in Figure 7, the longest line is that of index 2 and its neighbors according to the Y axis are the lines indexed 1, 3 and 4. Then for each neighboring line, a set of parameters is calculated including: o A difference in length between the considered line of the set and the neighboring line, o A distance between the center of the considered line and the center of the neighboring line, according to the abscissa axis X and the ordinate axis Y. These parameters are compared to predetermined thresholds. The distance threshold is preferably less than 25%, for example between 10 and 25%, for example equal to 15%. The distance threshold between the center of each line along the Y axis is preferably less than 50% of the length of the line considered (the line corresponding to a long pen), for example equal to 30% of the length of this line. The distance threshold between the center of each line along the X axis is preferably less than 50% of the length of the line considered, preferably less than 30%, for example equal to 25%. If these three parameters are below their respective thresholds, then the neighboring line is identified as corresponding to a long feather and added to the set.

[0086] The process is repeated until no more neighboring lines are added to the set of lines corresponding to long feathers.

[0087] In one embodiment, the other lines are automatically considered as lines corresponding to short feathers. However, for greater precision, the method comprises an identification 320 of the lines corresponding to short feathers. This identification is implemented, for each line corresponding to a long feather of the set formed previously, by following the position indices along the Y axis, that is to say by starting with the line located at the maximum vertical position of the set, and comprises: the identification, among the lines not belonging to the set of lines corresponding to long feathers, of the lines neighboring the line considered. In the example of figure 7, no neighbor would be identified for the line of index 1, but the neighboring line of index 3 would be identified for the line of index 2. For each line neighboring the line considered, we then calculate a difference in length between the line considered and the neighboring line, a distance along the vertical axis between the two lines, and a distance along the horizontal axis between a proximal or distal end of the line considered and a distal, respectively proximal, end of the neighboring line, according to the relative position on the X axis of the longest lines compared to the shortest lines.

[0088] The term "distal end" refers to the end furthest from the origin of the reference frame along the X axis, and the term "proximal end" refers to the closest end. If, as in the example of Figure 7, the shortest lines in the region obtained after processing 200 are located in the vicinity of the distal ends of the longest lines, then a distance is calculated between the distal end of the line in question and the proximal end of the neighboring line. If, conversely, the shortest lines are close to the proximal ends of the longest lines, a distance is calculated between the proximal end of the line in question and the distal end of the neighboring line. These parameters are compared to determined thresholds. The threshold for the difference in length is preferably between 50 and 80% of the length of the line considered, for example equal to 60%. The threshold in distance along the Y axis is preferably less than 50% of the length of the line considered, for example equal to 30%. The only threshold in distance along the X axis between the proximal end, respectively distal end of the line considered and the end distal, respectively proximal, of the neighboring line, is less than 50% of the length of the line considered, for example equal to 50%. If these three parameters are less than their respective thresholds, then the neighboring line is identified as corresponding to a short feather.

[0089] Once the lines corresponding to long feathers and the lines corresponding to short feathers are identified, the method may comprise a calculation 330 of a set of parameters from the region of interest resulting from the processing 200, these parameters then being provided to a trained classification model to determine the male or female sex of the chick.

[0090] In embodiments, these parameters include at least the number of lines corresponding to long pens and the number of lines corresponding to short pens.

[0091] Additionally, the parameters may further include an average angle between the lines and the horizontal, and an average gap or distance, measured vertically (Y axis) between two adjacent lines.

[0092] In embodiments, the parameters used for the model may further comprise one or more or any combination of all of the following parameters: minimum, maximum and / or average horizontal and / or vertical position of the centers of the lines, minimum, average and / or maximum distance, horizontally and / or vertically, between the center of two consecutive lines, minimum, average and / or maximum length of the lines corresponding to long feathers, minimum, average and / or maximum length of the lines corresponding to short feathers, average pixel intensity of a line corresponding to a short feather, average pixel intensity of a line corresponding to a long feather, difference in average pixel intensity between the lines corresponding to long feathers and the lines corresponding to short feathers, location relative to the horizontal axis of each line, a minimum, average and / or maximum angle between two lines, a minimum, average and / or maximum distance between the proximal ends of two lines corresponding to a long feather and a line corresponding to a short feather located between them, a minimum, average and / or maximum distance between the distal ends of two lines corresponding to a long feather and a line corresponding to a short feather located between them.

[0093] The parameters calculated at the end of step 300 are provided to a model trained to determine the sex of the chick, the model being trained so as to have two male / female output classes or alternatively, two classes comprising a first sex and an undetermined class.

[0094] The model used is, for example, but not limited to, a decision tree.

[0095] As indicated above, the model is trained on a database of annotated chick wing images. Preferably, the annotated images are images that have undergone the line extraction processing of step 200, and for which the steps of determining the lines corresponding to long feathers and the lines corresponding to short feathers, and of extracting the parameters 300 are also implemented, so that these parameters can be provided to the model for its training.

[0096] The annotation, that is to say the attribution to the region of interest considered of a male or female character of the chick, is carried out by an experienced operator according to the number of long feathers and short feathers visible in the region of interest considered.

[0097] This annotation may also include a degree of certainty associated with the determined sex, which may also be indicated by the operator. For example, on an image comprising 5 long feathers and 0 short feathers, the annotation may be “male; 100%”. In another example, on an image comprising 4 long feathers and 2 short feathers, the annotation may be “female, 100%”. The degree of certainty is preferably between 60 and 100% and in more uncertain cases, the annotation does not determine sex. In another third example, on an image comprising 3 long feathers and 1 short feather, the annotation may be “undetermined”. This type of image is then not retained for training the model.

[0098] In an embodiment in which several images are acquired for the same chick, for example at least 20 images, the processing described above can be implemented for each image. A result is thus obtained for each image concerning the sex of the chick, and the sex of the chick is then determined by a vote. majority, that is to say the result obtained most frequently, among the results obtained for all the images of the same chick.

[0099] Alternatively, the parameters extracted from each image can be provided to the trained model, and the determination of the sex of the chick is carried out by a majority vote,

[0100] Experimental results concerning the application of the method described above including the implementation of steps 200 and 300 and the application of a trained decision tree type model, on a parity population of 10,000 chicks, are reproduced below. The parameters extracted from the images and used in this experiment are: the number of long and short feathers, - the average angle between the lines and the horizontal, the average distance along the Y axis between two adjacent lines, and all of the additional parameters listed above in paragraph 92.

[0101] [Table 1]

Claims

DEPCT671. A method for determining the sex of chicks is performed by computer from images of chicks and consists of: - Examination (100) of the area of ​​interest in an image where the wing feathers are clearly visible; - Operation on the area of ​​interest, a sex determination model (400) trained on a training dataset consisting of images of the wings of male chicks and the wings of female chicks to determine whether the chicks are male or female.

2. The method according to claim 1 is applied to each image obtained from the same chick and consists of additional steps for determining the sex of chicks from the results obtained from the sex determination model for all images.3.One of the methods under the preceding claim in which the determination of the region of interest of an image (100) is performed includes: - scanning the image with a window of a specified size to divide the image into several regions; - for each region, the Haarfeature of that region (110) is calculated; - using each Haarfeature of the trained sexer (120) to determine whether the region is a feather; and - checking the region of interest of the image to be an area representing feathers.

4. One of the methods under claims 1 through 3 in which the additional processing of the region of interest (200) is performed to detect a set of lines corresponding to chick feathers in the image; checking (300) the set of parameters from the isolated lines; and applying (400) a sexer model to that set of parameters. 5.Claim 4. Processing of the region of interest (200) to determine a set of lines corresponding to hairs in the composite image includes: - processing the detection operation (210) of the region of interest and - in the resulting edge processing, using the linear edge finding technique (Houghtransform) to identify a set of lines (220) corresponding to visible hairs in the region of interest. Claim 6. Either one of Claim 4 or 5, in which parameter determination (300) of the isolated lines is performed includes: - identifying a set of lines corresponding to long hairs (310) and - identifying a set of lines corresponding to short hairs (320). Claim 7.The method based on the previous claim, which involves rotating (230) the area of ​​interest so that the lines representing the hairs extend almost horizontally, ordering each line in order of length and identifying a set of lines corresponding to the longest hairs (310), consists of: - Starting a set of lines corresponding to the longest hairs, such set contains the longest line. - The operation for each line included in such set is as follows: ๐ Identifying all adjacent lines of the line under consideration along the vertical axis. ๐ Calculating for each adjacent line the difference in length and distance between the centers of the adjacent lines. ๐ If the relative difference and distance are less than a given threshold, identify the adjacent line as the line corresponding to the longest hair and add a set of lines corresponding to the longest hairs. 8.The method for identifying the set of lines corresponding to short hairs (320) is comprised of the following steps for each line corresponding to the long hairs of the set, starting from the line at the highest vertical position of the set: - Identification among the lines not in the set of lines corresponding to the long hairs adjacent to the line being examined; - Calculation for each adjacent line of the difference in length, vertical distance between the line being examined and the adjacent line, and horizontal distance between the distal end of the line being examined and the distal end of the adjacent line; - If the calculated difference and distance are less than the relevant criteria, the adjacent line is identified as a line corresponding to a short hair.

9. Any one of the methods under claims 4 through 8, in which the parameters determined from the line are comprised of at least: - Number of lines corresponding to long hairs; - Number of lines corresponding to short hairs; - Average angle between the line and the horizontal; and - Average vertical deviation between two adjacent lines. 10.

11. One of the methods under Claims 4 through 9 in which a sex determination model is trained on a database of annotated training images, where each training image is obtained by using a procedure to define the region of interest and process the region of interest to examine a set of lines representing feathers and extract parameters from such lines and annotations, including an indication of the chick's sex and the associated confidence level, by checking the number of lines corresponding to long feathers and the number of lines corresponding to short feathers.

12. One of the methods under the preceding claim in which the method is applied to the same set of images of chicks and most often involves sex determination of the chick from the results obtained from the sex determination model.

13. A computer program product containing the code instructions for applying one of the methods under the preceding claim when executed by a processor.Equipment (1) for sex determination of chicks which includes at least: - a camera (20) adjusted to obtain at least one image of the chick and - a processor (10) configured to apply any of the methods of claims 1 through 11 on the images obtained by the camera.

14. Equipment (1) according to claim 13 which includes additionally a conveyor belt (30) which is modified to bring the chicks into the field of view of the camera (20) where the conveyor belt is modified to make the chicks unbalanced so that the chicks spread their wings when in front of the camera.15.Equipment (1) under either claim 13 or 14, which includes a conveyor belt (30), a first station for detecting first sex male or female chicks, which includes such camera (20) and a drive unit adjusted to pick up or eject from the conveyor belt where chicks are detected as first sex, where the processor (10) is configured to use the images obtained by the camera, which are the first sex classification model optimized for first sex detection, and the processor (10) is further configured to use the second sex classification model, which is improved for second sex detection on images obtained from chicks not detected as first sex;

Images