JP5090719B2 - Chick sex determination device and chick sex determination program - Google Patents

Description

  The present invention relates to a sex determination apparatus for chicks and a sex determination apparatus program for chicks.

  When raising chicks, the feed is different between males and females, so it is not efficient to feed the same chicks that have no distinction between males and females. In order to raise chicks efficiently, it is necessary to distinguish between males and females as soon as possible after birth. Currently, male and female determination of chicks is generally performed by human visual inspection, but it is easier if the male and female can be determined mechanically, and costs such as labor costs can be reduced. The merit of mechanization is great.

  Currently, there is a so-called wing determination method as an example of a sex determination method for chicks. The wing determination method uses the difference in wing characteristics between males and females in male wings, where male wings have a constant length, while female wings have alternating short and long wings. This is a method for determining the sex of chicks.

  An object of this invention is to provide the sex determination apparatus of a chick with a high determination rate, and the sex determination apparatus program of the chick which implement | achieves it.

  As one means for solving the above-described problem, a sex determination apparatus according to the present invention includes a preprocessing unit that performs preprocessing on image data, and feature amount extraction that extracts feature data based on the preprocessed image data. A processing unit, and a determination processing unit that determines sex based on the extracted feature data.

  Moreover, in this means, although not limited, it is preferable that the preprocessing unit normalizes the luminance value of the pixel in the image data.

  Moreover, in this means, although not limited, the feature amount extraction processing unit performs azimuth filtering on the preprocessed image data, and 2 of the luminance value of the pixel in the preprocessed image data. Performing binarization, binarizing the luminance value of a pixel in the preprocessed image data, extracting a portion corresponding to a wing, performing wavelet transform on the preprocessed image data, It is preferable to perform at least one of the following.

  Further, in this means, although not limited, when wavelet transform is performed on the preprocessed image data, high-pass filtering is performed on the image data subjected to wavelet transform, and multiplier processing is performed. It is preferable to perform at least one of the following. In addition, when performing a multiplier process, it is more preferable to perform a differential process, and it is still more preferable to perform differential filtering.

  In this means, although not limited, the feature amount extraction processing unit binarizes the luminance value of the pixel in the preprocessed image data and extracts a portion corresponding to the wing. , Performing thinning on the part corresponding to the wing, combining the part corresponding to the wing, rotating the image data, and further, between the vertices of the part corresponding to the wing It is preferable to perform at least one of obtaining at least one of the distance and the length.

  Moreover, in this means, although not limited, it is preferable that the determination processing unit uses at least one of a Bayes identification method or a linear determination method.

  In addition, as another means for solving the above-described problem, a chick sexing program according to the present invention includes a preprocessing unit that preprocesses image data on a computer, and the preprocessed image data. A feature amount extraction processing unit that extracts feature amount data based on the extracted feature amount data and a determination processing unit that determines sex based on the extracted feature amount data.

  Moreover, in this means, although not necessarily limited, it is also preferable to function as a preprocessing unit that normalizes the luminance value of the pixel in the image data.

  Further, in this means, although not limited, a feature extraction processing unit that performs azimuth filtering on the preprocessed image data, and binarization of pixel luminance values in the preprocessed image data. A feature extraction processing unit, a feature extraction processing unit that binarizes luminance values of pixels in the preprocessed image data, a feature extraction processing unit that extracts a portion corresponding to a wing, and the preprocessed image data It is also preferable to function as at least one of a feature extraction processing unit that performs wavelet transform.

  Further, in this means, although not limited thereto, if wavelet transform is performed on the preprocessed image data, a feature extraction processing unit that performs high-pass filtering on the image data subjected to wavelet transform It is also preferable to function as at least one of a feature extraction processing unit that performs multiplier processing. In addition, when a multiplier process is performed, it is more preferable to function as a feature extraction processing unit that performs differential processing, and it is even more preferable to function as a feature extraction processing unit that performs differential filtering.

  Further, in this means, although not limited, the feature amount extraction processing unit binarizes the luminance value of the pixel in the preprocessed image data, and extracts a portion corresponding to the wing. When functioning as an extraction processing unit, a feature extraction processing unit that performs thinning on the portion corresponding to the wing, a feature extraction processing unit that performs combination processing on the portion corresponding to the wing, and a feature that performs rotation processing on image data It is also preferable to function as at least one of an extraction processing unit and, further, a feature extraction processing unit that obtains at least one of a distance or a length between vertices of a portion corresponding to a wing.

  Moreover, although not limited, it is also preferable that this means function as a determination processing unit that uses at least one of the Bayes identification method or the linear discrimination method.

  Thus, the present invention provides a sex determination apparatus for chicks with a higher determination rate and a sex determination apparatus program for chicks that realizes the determination ratio.

  Embodiments of the present invention will be described below with reference to the drawings. However, the present invention can be implemented in many different modes and is not limited to the embodiments shown below. Note that in this specification, portions having the same or similar functions are denoted by the same reference numerals, and repeated description thereof is omitted.

  FIG. 1 shows functional blocks of a chick sex determination apparatus (hereinafter also referred to as “the present sex determination apparatus”) according to the present embodiment. As shown in the figure, the sex determination apparatus 1 includes at least a preprocessing unit 2, a feature extraction processing unit 3, and a determination processing unit 4. The sex determination apparatus 1 is not limited, but can be realized by executing a program (chicken sex determination program) stored in a recording medium such as a hard disk of a personal computer, for example. In the present embodiment, description will be made using an example realized by executing a program. Note that the user of this device needs to capture and capture in advance an image of the wings of a chick by an imaging device such as a camera prior to the sex determination by the sex determination device.

  The pre-processing unit 2 can perform a process of adjusting an influence generated in photographing such as illumination prior to the process in the acquired image of the wing part. In the present embodiment, normalization of the luminance value in the image data (hereinafter simply referred to as “image data”) of the wing part of the chicken captured in advance is performed. Thereby, the influence of the illumination used when imaging a wing | blade part can be reduced. FIG. 2A shows a photograph of the male wing portion, and FIG. 2B shows a photograph of the female wing portion.

The normalization of the luminance value is not limited, but the wings are emphasized by adjusting the histogram in the image data. Although the histogram conversion is not limited, it is preferable to perform the histogram conversion using an equation represented by the following equation (1). In the following formula, x represents the luminance value of a pixel in image data before processing (hereinafter referred to as “original image data”), and x _max and x _min are the maximum luminance value and the minimum luminance value in the original image data, respectively. Indicates the value. The y _s denotes the luminance value of the pixel in the processed image data (hereinafter referred to as "luminance value normalized image data".). Y _max is 255.

The feature extraction processing unit 2 can extract a portion corresponding to a wing and extract the feature amount. Although not limited, specifically, the luminance value of the pixel in the luminance value normalized image data is binarized and extracted as a portion corresponding to a wing. Here, binarization means a process of assigning 0 if the luminance value of a certain pixel in the image data is smaller than the threshold value Th, and assigning the maximum value (255) if it is larger. The binarization is specifically expressed by the following formula (2). In the following, f (x, y) represents the luminance value in the normalized luminance value image data, and g (x, y) represents the image data after binarization processing (hereinafter “after binarization”). "Image data"). FIG. 3 shows a photograph of an example of binarized image data. By binarization, it is effective to separate wings caused by uneven light reflection due to unevenness existing on the surface of the wings (a single wing would be recognized as multiple wings) Can be suppressed.

However, as shown in FIG. 3, even when binarization is performed, the separation of the wings may not be completely suppressed. Therefore, in this embodiment, attention is paid to the wing part separated by binarization, and the wing part is combined in consideration of the angle. FIG. 4 shows an example of the wing A combining process. The coupling process is not limited, but as shown in FIG. 4, first, an angle θ _A formed by a certain reference axis (horizontal axis in FIG. 4) and the axis of the wing A is obtained. On the other hand, a straight line connecting one end point of the wing A and one end point of the wings B, C, D is obtained, and angles θ ′ _B , θ ′ _C , θ ′ _D formed by the respective straight lines and the reference axis are obtained. Then, it is determined whether or not any of θ ′ _B , θ ′ _C , and θ ′ _D has an angle close to θ _A, and wings having close angles are combined. In the example of FIG. 4, since the angles of the wing A and the wing B are substantially equal, it is determined that the wing A and the wing B are one wing, and the combining process is performed.

  Here, the method of determining the axis of each wing is not limited, but it is preferable to determine it by performing a thinning process. More specifically, when a portion extracted as a wing has a width, it is preferable to express this width with one line. FIG. 5 shows image data after this processing (hereinafter referred to as “image data after thinning”). 5A shows image data before luminance value normalization, FIG. 5B shows image data after luminance value normalization, and FIG. 5C shows image data after thinning. In addition, after the thinning process, it is also preferable to perform a straight line approximation process for approximating the thin line to a straight line in order to perform the combining process more easily.

In the feature extraction processing unit 3, it is preferable to perform rotation processing on the image data and adjust the direction of the wing in order to easily extract the change in the tip of the wing as a feature amount. The rotation angle in this rotation process is not limited, but is preferably the average value φ _ave of the angle θ of each wing axis with respect to the reference axis obtained in the above combination process. The rotation processing is not limited, but a certain point, for example, a pixel at the position of (x ₁ , y ₁ ) and a pixel at the position of (x _n , y _n ) (n is the x direction and When the center point in the y direction indicates the number thereof is set as the rotation center, it is preferable to process by the following formula (3). FIG. 6 is a diagram for explaining this rotation model.

Then, the feature extraction process 3 extracts tip change information as feature amount data used for determination from the image data. More specifically, the positions of the vertices of the wings are obtained for each and derived as a difference value (hereinafter referred to as “difference derivation process”). Specifically, the coordinates of each vertex of the wing are [x ₁ , y ₁ ], [x ₂ , y ₂ ],... [X _N , y _N ] (N is the number of wings), and adjacent to each vertex. A difference value Δy of the matching y coordinates is derived, and an average value m (Δy) is obtained. Specifically, the following equations (4) and (5) are used. Incidentally, i = 1, 2, 3 in the formula (5), a ... N-1, _{l n} denotes the length of each blade. In this embodiment, since the rotation process is performed, the x coordinate is not considered, but it is naturally possible to set the distance between coordinates including the x coordinate as well as the difference value of the y coordinate. . In addition, regarding the difference value Δy, it is preferable to obtain the variance var (Δy) from the viewpoint of making the determination more accurate (see, for example, Expression (6)).

  The determination processing unit 4 according to the present embodiment performs sex determination. Although it is not limited as long as sex determination can be performed, it is preferable to use a Bayes identification method. Here, the “Bayes identification method” is a method of determining, based on the probability, which class the feature quantity to be identified belongs to from the distribution of the feature quantity of each class obtained in advance. Bayesian identification is divided into two steps: learning and recognition.

In the learning step in the Bayes identification method, probability density functions P (x | ω ₁ ) and P (x | ω ₂ ) representing the distribution of feature quantities of male and female are obtained. Note that the probability density functions P (x | ω ₁ ) and P (x | ω ₂ ) represent the occurrence probabilities of x belonging to the class ωi. Here, x represents a feature amount used for the determination process, ω ₁ represents a male class, and ω ₁ represents a female class.

On the other hand, in the recognition step in the Bayes identification method, feature quantities are extracted from the chickens whose sex is to be determined, and posterior probabilities P (ω ₁ | x), P (ω ₂ | x) chickens are extracted based on the feature quantities. Determine whether is a male or female. Here, the posterior probability can be obtained using the following equation (7). In the following formula (7), P (ω ₁ ) and P (ω ₂ ) are the occurrence probabilities of males and females, respectively. P (ω ₁ ) = P (ω ₂ ) = 0.5 assuming that the occurrence probabilities of male and female are equal. The male / female determination process is performed based on the following equation (8).

  As described above, according to the present embodiment, it is possible to provide a sex determination apparatus for chicks with a high determination rate and a sex determination program for chicks that realizes the apparatus.

(Embodiment 2)
The embodiment described above is useful in that it can provide a sex determination apparatus for chicks with a high determination rate and a sex determination program for chicks that realizes it. Down is included, and depending on the part to be photographed, in the process of extracting the wing part, the down is recognized as a wing and there is no possibility of affecting the determination rate. In the above embodiment, the wings are represented by straight lines, and the wings are combined based on the angle. Therefore, when the wings are bent, the angle between the separated wings is different. It is not without the case that it brings difficulties to the connection. Therefore, in this embodiment, a male determination device that can maintain a high determination rate even when down is included, and a sex determination program for chicks that realizes the male determination device are realized. Although the present embodiment will be described below, points different from the first embodiment will be mainly described, and the description of the same parts as those of the first embodiment will be omitted unless mentioned.

  FIG. 7 shows a functional block diagram of the sex determination device according to the present embodiment. Similar to the first embodiment, the sex determination apparatus 1 according to the present embodiment includes at least a preprocessing unit 2, a feature extraction processing unit 3, and a determination processing unit 4.

  One of the features of the feature extraction processing unit 3 in this embodiment is to perform azimuth filtering that emphasizes a straight line. In general, chicken wings grow in a straight line, so that even when down is included by performing azimuth filtering that emphasizes the straight line, the wings can be emphasized more effectively.

In azimuth filtering, the classification of image data and the number of azimuth filters can be adjusted as appropriate and are not limited. However, an example in which the image data is divided into 7 × 7 and the number of azimuth filters is nine. As shown in FIG. Further, the azimuth filtering process is performed according to the following equations (9) and (10). In the following formulas (9) and (10), f (x, y) represents the luminance value of the pixel of the image data after normalization of the luminance value, and h _i (x, y) represents the orientation filter. The filtering coefficient in the pixel where is located, and gi (x, y) represents the luminance value of the pixel of the image data after performing the azimuth filtering. In the following formula, * indicates a convolution operation. i indicates the number of orientation filters, and in this case, i = 1, 2,... FIG. 9 shows a diagram before and after orientation filtering. FIG. 9A shows image data before azimuth filtering, and FIG. 9B shows image data after azimuth filtering.

  The feature extraction unit 3 is capable of extracting a portion corresponding to a wing and extracting the feature amount. The feature extraction unit 3 binarizes, thins, combines, and rotates a luminance value of a pixel in image data. The difference derivation process can be the same as that of the embodiment.

In addition to the above, in the present embodiment, in addition to the Δy difference value deriving process (for example, Equations (4) to (6)) of the first embodiment, it is preferable to perform the Δl difference value deriving process. In addition to the difference value of Δl, it is preferable to obtain the average and variance of l. The equation for calculating the difference value Δl is shown in the following equation (11), the equation for calculating the average of Δl is shown in the following equation (12), and the equation for calculating the variance of Δl is shown in the following equation (13).

The determination processing unit 4 can use the same method as in the first embodiment, but in this embodiment, a linear discrimination method is used. The linear discriminating method can be performed according to the formula shown by the following formula (14), although not limited thereto. Note that x represents a change in the tip, which is feature quantity data, ω1 represents a male, ω2 represents a female class, and Th _linear represents a threshold used in the linear discrimination method.

  As described above, according to the present embodiment, in addition to the effects of the first embodiment, a high determination rate can be maintained even when down is included.

(Embodiment 3)
In the second embodiment, even if the image data includes a portion corresponding to down, a hen's sex determination apparatus with a higher determination rate, and a chick's sex determination program for realizing the device are provided. This is useful in that it can be removed, but the effect of this down can be removed more efficiently. Although the present embodiment will be described below, differences from the second embodiment will be mainly described, and the description of the same parts as those of the second embodiment will be omitted unless mentioned.

  FIG. 10 shows a functional block diagram of the sex determination apparatus for chicks according to this embodiment. Similar to the first embodiment, the sex determination apparatus 1 according to the present embodiment includes at least a preprocessing unit 2, a feature extraction processing unit 3, and a determination processing unit 4.

  The feature amount extraction processing unit 2 according to the present embodiment performs wavelet transform. In general, the down includes a dense image and an isolated image. The dense down appears to be flat in the image and is considered to be composed of low frequency components. On the other hand, since the wings are elongated and arranged at regular intervals, it is considered that they are composed of components having a higher frequency than dense down. Therefore, dense down and wings can be first separated by performing wavelet transform to emphasize regions having different frequency components. The wavelet transform is not limited as long as frequency components can be separated, but it is preferable to use, for example, a fast Fourier transform (hereinafter referred to as “FFT”). FIG. 11A shows an image before the fast Fourier transform, and FIG. 11B shows a spatial frequency image obtained after the fast Fourier transform.

Next, the feature amount extraction processing unit 2 performs high-pass filtering using a high-pass filter. By performing high-pass filtering, low frequency components corresponding to dense down can be removed, and high frequency components corresponding to the wings can be efficiently left. Here, the high-pass filter is not limited, but a circular filter having a radius R is preferably used. By making the radius R correspond to the wing frequency, low frequency components corresponding to dense down can be removed. When the width of the wing in the image is approximately B pixels, the average period representing the wing can be considered as 2B pixels, and the radius R of the circular filter can be expressed by the following equation (15). . In the following formula, width indicates the width of the image. FIG. 12A shows an image before FFT return, and FIG. 12B shows an image after high-pass filter processing and inverse FFT.

Moreover, although not necessarily limited, it is preferable that the feature quantity extraction processing unit 3 removes not only downs that are dense but also isolated downs. Utilizing the fact that the wing spectrum is strong while the isolated down spectrum is weak, multiply the absolute value by the k-th power to make the strong spectrum relatively strong, and the weak spectrum relatively weakened ( Hereinafter, “multiplier processing” is preferable. This process can be realized by the following equation (16). In this equation, g (x, y) is the luminance value of the pixel in the image data after enhancement, F is FFT, F-1 is inverse FFT, and k is a constant. FIG. 13A shows an image before the multiplier process, and FIG. 13B shows an image after the multiplier process (k = 0.5).

  Moreover, after performing this multiplier processing, it is preferable to remove the portion corresponding to down by cutting out the portion corresponding to the wing from the image. Depending on the conditions of the multiplier processing, the luminance value of the pixels other than the part corresponding to the wings and down (hereinafter referred to as “background part”) may increase, so binarization alone may not be sufficient. . However, as described above, since the background portion is flat, the change in the pixel value is smooth, while in the wing portion, since the wings are lined up, the change in the luminance value of the pixel is considered to be abrupt. Therefore, if the image data after the multiplier process is differentiated to obtain the image data after the differentiation process, it is considered that the differential value in the background portion is small and the differential value in the portion corresponding to the wing is large. Therefore, if differential filtering is performed on the image data after this differential processing using a differential filter, a portion corresponding to a wing can be extracted efficiently. Thereafter, by performing the above binarization, the subsequent processing becomes more efficient. FIG. 14 shows an example of the differential filter.

  In addition, the feature quantity extraction processing unit 3 according to the present embodiment is also characterized in that the combination processing can be performed with high accuracy and high efficiency even when the wing is bent. The combining process here will be described with reference to FIGS. 15 and 16.

  FIG. 15 is a diagram illustrating a model expressing a wing as a three-dimensional shape. When a wing is cut along a plane perpendicular to the long axis of the wing (a plane including the end axis), the cut end has a mountain shape. The integral value in the image data after the differential processing described above changes from positive to negative at the peak of the mountain. The height of the mountain varies depending on the lighting and the unevenness of the wings, but there are mountains where there are wings. Therefore, if a ridge line is created by tracking the point where the differential value existing in the area corresponding to a certain wing changes from positive to negative, and the ridge line reaches another separated wing, It is possible to perform the combining process as if there were. FIG. 16 shows an example of combining by creating a ridge line. By doing in this way, even if it is a case where a wing is bent, a joint process can be performed. In the subsequent thinning process, the same one as in the above embodiment can be adopted. In particular, unlike the above-described embodiment, this method requires parameter setting for examining the angle between the wings and performing coupling, but using the proposed method eliminates the need for parameter setting for that purpose. There is also an effect.

  Note that the feature quantity extraction processing unit 3 performs a rotation process to extract feature quantity data such as a difference value Δy. In addition to the method described in the above embodiment, the feature quantity extraction process unit 3 uses a wavelet transform (for example, FFT) for the rotation process. It is also preferable to perform a principal component analysis from the properties of the frequency image obtained at the time of obtaining the rotation angle. As a property of the frequency image, when a two-dimensional wavelet transform (for example, FFT) is performed on a striped pattern in which bright and dark lines are arranged at regular intervals, the spatial frequency image has a pixel distribution length in a direction orthogonal to the stripe. The axis appears. Since the wings are arranged at regular intervals and have an elongated shape, they can be regarded as striped. That is, the direction of the wing is a direction that is orthogonal to the pixel distribution of the spatial frequency image. Therefore, the major axis and the minor axis of the spatial frequency image distribution can be obtained by performing principal component analysis. Note that the first eigenvector obtained by performing the principal component analysis has a direction orthogonal to the wing, and the second eigenvector has a direction parallel to the wing. Therefore, the rotation angle can be determined based on this angle.

  As mentioned above, according to this embodiment, in addition to the effect in the said embodiment, the influence by a down can be removed more efficiently.

  A computer experiment was actually conducted on the above embodiment to confirm its performance. This will be described below.

  This example is the result of a computer experiment for the above embodiment. In this embodiment, the number of images used is 196 males and 196 females, and the number of pixels in one image data is 256 × 256. The total number of images used was 392, but the Leave-one-out method was used to increase the accuracy of the computer experiment. In this example, the threshold for binarization was set to 160, and the coupling angle was set to 30 degrees or less.

FIG. 17 shows the distribution of the feature data obtained as a result of the computer experiment for the first embodiment, FIG. 18 shows the distribution of the feature data obtained as a result of the computer experiment for the second embodiment, and FIG. The distribution of the feature data obtained as a result of the computer experiment is shown respectively. In addition, in the result of the computer experiment for the second embodiment and the result of the computer experiment for the third embodiment, the Bayes identification method similar to that of the first embodiment is used for the uniformity of the determination rate. FIG. 20 shows the probability density function used for the first embodiment, FIG. 21 shows the probability density function used for the second embodiment, and FIG. 22 shows the probability density function used for the third embodiment. . As a comparison with these, as a result of a computer experiment in a case where wings are ideally extracted (hereinafter referred to as “ideal case”), probability density functions used for the results are shown in FIGS. 23 and 24, respectively. . Here, “when the wings are ideally extracted” refers to a case where all the wings are completely extracted without erroneous extraction of the down and separation of the wings. In addition, these determination rates are collectively shown in Table 1 below.

  In any of the above results, a high determination rate could be obtained. In particular, the azimuth filtering is mainly different between the first embodiment and the second embodiment, and it can be considered that this difference contributes to the determination rate, and the excellent effect of the azimuth filtering can be confirmed. In the third embodiment, down is more efficiently removed, and a differentiation process is used in the combining process, so that a higher determination rate can be obtained.

  The present invention has industrial applicability as a sex determination apparatus and sex determination program for chicks.

It is a figure which shows the functional block of the sex determination apparatus of the chicken which concerns on Embodiment 1. FIG. It is a figure of the wing part of a chick. It is a figure which shows the image data after binarization. It is a figure which shows the example of a joint process of a wing | blade. It is a figure for demonstrating a thinning process. It is a figure explaining the model of rotation of an image. It is a figure which shows the functional block of the sex determination apparatus of the chicken which concerns on Embodiment 2. FIG. It is a figure which shows the example of a direction filter. It is a figure for demonstrating a direction filtering process. It is a figure which shows the functional block of the sex determination apparatus of the chick which concerns on Embodiment 3. FIG. It is a figure which shows the result of having performed fast Fourier transform to image data. It is a figure for demonstrating a high pass filter process. It is a figure for demonstrating a multiplier process. The figure which shows the example of a differential filter. It is a figure which shows the model which expressed the feather | wing as a three-dimensional shape. It is a figure for demonstrating the combination process of a wing | blade. It is a figure which shows distribution of the differential quantity data calculated | required as a result of the computer experiment with respect to Embodiment 1. FIG. It is a figure which shows distribution of the differential quantity data calculated | required as a result of the computer experiment with respect to Embodiment 2. FIG. It is a figure which shows distribution of the differential amount data calculated | required as a result of the computer experiment with respect to Embodiment 3. FIG. It is a figure which shows the probability density distribution function calculated | required in Embodiment 1. FIG. It is a figure which shows the probability density distribution function calculated | required in Embodiment 2. FIG. It is a figure which shows the probability density distribution function calculated | required in Embodiment 3. It is a figure which shows the probability density function of the result of a computer experiment in case a wing | blade is extracted ideally. It is a figure which shows the probability density function of the result of a computer experiment in case a wing | blade is extracted ideally.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Sex determination apparatus, 2 ... Pre-processing part, 3 ... Feature length extraction process part, 4 ... Determination process part

Claims (16)

A preprocessing unit for performing preprocessing on image data;
A feature amount extraction processing unit for extracting feature amount data based on the preprocessed image data;
A determination processing unit that determines sex based on the extracted feature data ,
The feature amount extraction processing unit performs azimuth filtering processing on the preprocessed image data, binarizes pixel luminance values, extracts a plurality of portions corresponding to wings, and corresponds to the wings A sex determination apparatus for chicks that performs thinning processing, straight line approximation processing, and combining processing on each of the portions, finds the vertices of the portions corresponding to the wings, and finds the distance between the vertices.   The sex determination apparatus for chicks according to claim 1, wherein the feature quantity extraction processing unit performs wavelet transform, high-pass filtering, and inverse wavelet transform on the preprocessed image data. The sex determination apparatus for chicks according to claim 2, wherein the feature amount extraction processing unit performs a multiplier process on the image data after the wavelet transform .   4. The chick sex determination apparatus according to claim 3, wherein the feature extraction unit performs differential processing and differential filtering on the image data. The sex determination apparatus according to claim 3 , wherein the feature amount extraction processing unit performs a rotation process on the image data. The sex determination apparatus according to claim 3, wherein the feature amount extraction processing unit obtains a length of a portion corresponding to the wing.   The sex determination apparatus according to claim 1, wherein the determination processing unit uses a Bayes identification method.   The sex determination apparatus according to claim 1, wherein the determination processing unit uses a linear discrimination method. On the computer,
A preprocessing unit for performing preprocessing on image data;
A feature amount extraction processing unit for extracting feature amount data based on the preprocessed image data;
A determination processing unit for determining the sex based on the extracted feature data, and a sex determination program for chicks to function .
The feature extraction processing unit performs azimuth filtering processing on the preprocessed image data, binarizes a luminance value of a pixel, extracts a plurality of portions corresponding to wings, and a portion corresponding to the wings A program for determining the sex of chicks that performs thinning processing, straight line approximation processing, and combination processing on each of the above, and obtains the vertices of the portion corresponding to the wing, and obtains the distance between the vertices. On the computer,
10. The sex determination program for chickens according to claim 9 , wherein the pre-processed image data is functioned as a feature quantity extraction processing unit for performing wavelet transform, performing high-pass filtering, and performing inverse wavelet transform . On the computer,
The sex determination program for chicks according to claim 9, for causing the image data after the wavelet transform to be performed to function as a feature amount extraction processing unit for performing a multiplier process on the image data . On the computer,
10. The sex determination program for chicks according to claim 9, for causing the image data to function as a feature amount extraction processing unit that performs differential processing and differential filtering on the image data. On the computer,
The sex determination program according to claim 9 for functioning as a feature amount extraction processing unit that performs a rotation process on the image data. On the computer,
The sex determination program according to claim 9 for functioning also as a feature amount extraction processing unit for obtaining a length of a portion corresponding to the wing. On the computer,
The sex determination program according to claim 9 for functioning also as a determination processing unit using a Bayes identification method. On the computer,
The sex determination program according to claim 9 for functioning also as a determination processing unit using a linear discriminant method.