Chicken head feature determination method based on 360-degree panoramic scanning

Abstract

The invention provides a method for measuring chicken head features based on 360-degree panoramic scanning. The method comprises the following steps: acquiring an internal scene image of a chicken house by adopting a depth camera array; performing color correction on the internal scene image based on a standard colorimetric card; performing chicken head three-dimensional modeling in the chicken house based on the internal scene image after color correction to obtain a chicken head three-dimensional model; performing cockscomb feature extraction based on the chicken head three-dimensional model to obtain a cockscomb feature image; according to the method, all-around information collection of chicken flocks in a chicken house is achieved through 360-degree panoramic detection, farmers can know the growth condition and the nutrition condition of the chickens more comprehensively, the YOLO model and the U-Net model are combined to complete image segmentation, and the image segmentation efficiency is improved. The efficiency and accuracy of image annotation and segmentation are further improved, and meanwhile, image areas needing to be processed can be greatly reduced, so that the calculation cost is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of image processing and three-dimensional modeling, and in particular to a method for measuring chicken head features based on 360-degree panoramic scanning. Background Art

[0002] The comb is one of the important appearance characteristics of chickens. It is closely related to the age, gender, health status and reproductive ability of chickens. By observing and evaluating the comb phenotype, the health status, growth and development, and reproductive potential of chickens can be evaluated. The phenotypic characteristics of the comb include size, color and shape. In the breeding of breeder chickens, the development of the comb is closely related to the growth of individual chickens, especially the comb of roosters will be significantly enlarged and bright red in the sexual maturity stage, so the development of the comb is an important indicator to measure the sexual maturity of roosters; some studies have also pointed out that the size and shape of the comb are also related to the semen quality of roosters and the egg-laying performance of hens. For example, in the study of Nandan Yao chickens, it was found that the semen volume of roosters with large combs was higher than that of roosters with small combs, and the size and color of the combs and wattles of roosters in this cross cage determined their attractiveness to hens; in the study of Roman gray laying hens, it was found that the age of hens in laying was extremely significantly negatively correlated with the comb height and comb length, and the egg production was extremely significantly positively correlated with the comb length and comb height. A healthy rooster's comb is upright, bright red, plump, and has soft and smooth tissue. In production, bright red combs of a group of roosters are an important sign of good ventilation in the chicken house. The phenotype of an individual rooster's comb is closely related to the health of the chicken. Dark red combs are seen in acute infectious diseases such as avian influenza, Newcastle disease, cholera, and fowl typhoid. Yellow or white combs indicate that the chicken has anemia or fatty liver causing liver bleeding, splenomegaly and bleeding, coccidiosis, or severe parasitic infection. Pox spots on the comb and bristles indicate fowl pox, etc.

[0003] It can be seen that in chicken breeding, accurate evaluation of comb phenotype is one of the important means to judge the sexual development and reproductive potential of breeder chickens, which helps to improve breeding efficiency; in the chicken production process, rapid and accurate evaluation of comb phenotype is one of the important ways to judge the health status of individuals. By regularly testing and recording the characteristics of combs, breeders can timely discover and deal with health problems, reduce the risk of disease transmission, and improve the survival rate and production efficiency of chickens. Traditional comb detection usually relies on manual observation, which is not only time-consuming and labor-intensive, but also easily affected by subjective factors. Although there are some means to use image recognition to realize image recognition of combs, due to the internal structure of the chicken house, it is impossible to obtain a full-angle image of the chicken flock, and it is difficult to obtain the characteristic information of the comb in all directions. A fixed image acquisition platform is often set up, and the chicken needs to be fixed on a specific device and take pictures using a fixed camera position, which is time-consuming and labor-intensive, and will cause greater stress to the chicken. Summary of the invention

[0004] In view of this, the present invention proposes a method for measuring chicken head features based on 360-degree panoramic scanning to solve the problems existing in the above-mentioned prior art, realize automatic detection of chickens, and at the same time improve detection efficiency and accuracy and reduce labor costs.

[0005] To achieve the above object, the present invention proposes a method for measuring chicken head features based on 360-degree panoramic scanning, comprising the following steps:

[0006] A depth camera array is used to obtain images of the internal scenes of the chicken house;

[0007] Performing color correction on the internal scene image based on a standard colorimetric card;

[0008] Perform three-dimensional modeling of the chicken head in the chicken house based on the color-corrected internal scene image to obtain a three-dimensional model of the chicken head;

[0009] Extracting comb features based on the three-dimensional model of the chicken head to obtain a comb feature image;

[0010] The comb measurement information is obtained based on the chicken head three-dimensional model and the comb characteristic image.

[0011] Optionally, the process of acquiring an internal scene image of a chicken house using a depth camera array includes:

[0012] Acquire the imaging range of the depth camera array;

[0013] Setting a depth camera array position based on the imaging range;

[0014] The depth camera array is arranged according to the position, and internal scene images of several chicken houses at different angles are acquired based on the arranged depth camera array.

[0015] Optionally, the process of setting the position of the depth camera array based on the imaging range includes:

[0016] Set up a camera rail array around the top of the chicken house;

[0017] Get the locations of the four corners and the center of the top of the chicken house;

[0018] The installation position of the camera rail array is determined according to the positions of the four corners and the center of the top of the chicken house and the imaging range of the depth camera array. After the camera rail array is installed, a lifting bracket is installed on the camera rail array, and the depth camera array is set on the lifting bracket.

[0019] Optionally, in the process of acquiring the internal scene images of the chicken house using several depth camera arrays, the light environment in the chicken house is detected and the process of acquiring the images by the depth camera array is supplemented with light based on the detection results.

[0020] Optionally, detecting the light environment in the chicken house and supplementing the light for the process of acquiring the image by the depth camera array based on the detection result includes:

[0021] The light environment signal in the chicken house is received and converted to obtain a light environment control digital signal of the chicken house. Based on the light environment control digital signal, an infrared fill light device is used to fill light in the process of acquiring images by the depth camera array.

[0022] Optionally, the step of converting the light environment signal to obtain a light environment control digital signal for the chicken house includes:

[0023] Performing nonlinear conversion on the light environment signal to obtain a light environment control electrical signal;

[0024] The light environment control electrical signal is converted into a light environment control digital signal through a data acquisition card.

[0025] Optionally, the process of performing color correction on the internal scene image based on a standard colorimetric card includes:

[0026] A mapping relationship matrix between the RGB values ​​of the standard colorimetric card and the RGB values ​​of the internal scene image is constructed by using polynomial regression. After the mapping relationship matrix is ​​solved by the least squares method, the RGB values ​​of the internal scene image are substituted into the mapping relationship matrix, color correction is performed based on the RGB values ​​of the standard colorimetric card, and the corrected RGB values ​​of the internal scene image are output.

[0027] Optionally, the process of performing three-dimensional modeling of a chicken head in a chicken house based on the color-corrected internal scene image includes:

[0028] The chicken head in the internal scene image is identified based on the image segmentation model and the image detection model in the deep learning network, and the point clouds of the chicken head at different angles are registered based on the point cloud registration algorithm, and the chicken head is three-dimensionally reconstructed to obtain the three-dimensional model of the chicken head.

[0029] Optionally, the process of extracting the comb features based on the chicken head three-dimensional model includes:

[0030] After smoothing and noise removal of the chicken head three-dimensional model, the processed chicken head three-dimensional model is automatically segmented using a Reeb graph to obtain a number of feature blocks, which are statistically analyzed, and feature blocks belonging to the comb are obtained based on the statistical results, and the comb feature image is obtained based on the feature blocks belonging to the comb.

[0031] Optionally, the process of obtaining the comb measurement information based on the chicken head three-dimensional model and the comb characteristic image includes:

[0032] The shape of the comb feature block is used as the shape of the comb, and the color of the comb feature image is used as the color of the comb;

[0033] Based on the internal scene image, obtain a reference object in the chicken house and obtain the actual size information of the reference object, perform feature extraction on the outline of the reference object based on OpenCV, obtain the pixel size of the reference object in the internal scene image, obtain the ratio coefficient between the pixel size of the reference object and the actual size, and obtain the actual size of the comb based on the ratio coefficient and the pixel value size of the comb feature image;

[0034] The comb measurement information is constructed based on the shape, color and actual size of the comb.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention adopts a depth camera array and realizes 360° scene detection in the chicken house by setting slide rails and lifting brackets in the chicken house, which can realize all-round information collection of the chickens in the chicken house, helping farmers to have a more comprehensive understanding of the growth and nutritional status of the chickens.

[0037] By performing color correction on the image, the present invention can ensure that the color, skin condition, comb size and other characteristics of the chickens in the image are presented more realistically and accurately, helping breeding technicians, veterinarians or farmers to more accurately judge the health status of the chickens, and more accurately identify the number, location, behavior and other information of the chickens, so as to take timely treatment measures.

[0038] The present invention controls the light in the chicken house, which helps the depth camera array to obtain clearer images of the chicken house environment and provide farmers with more accurate and reliable data support.

[0039] The present invention combines the YOLO model and the U-Net model to complete image segmentation, which can further improve the efficiency and accuracy of image annotation and segmentation, and at the same time can greatly reduce the image area that needs to be processed, thereby reducing the computing cost, which is of great significance for processing large-scale data sets or real-time application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only used for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. In the accompanying drawings:

[0041] Figure 1 This is a flow chart of a method for measuring chicken head features based on 360-degree panoramic scanning in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to be able to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] This embodiment proposes a method for measuring chicken head features based on 360-degree panoramic scanning, such as Figure 1 As shown, the following steps are included:

[0044] S1. Use a depth camera array to obtain an image of the interior scene of the chicken house;

[0045] S2. Color correction of the internal scene image based on a standard colorimetric card;

[0046] S3. Perform three-dimensional modeling of the chicken head in the chicken house based on the color-corrected internal scene image to obtain a three-dimensional model of the chicken head;

[0047] S4. Extracting the comb features based on the three-dimensional model of the chicken head to obtain a comb feature image;

[0048] S5. Acquire comb measurement information based on the chicken head three-dimensional model and the comb feature image.

[0049] As a preferred embodiment, the process of step S1. using a depth camera array to acquire an internal scene image of a chicken house further comprises the following steps:

[0050] S1.1 obtain the imaging range of the depth camera array;

[0051] S1.2 setting the position of the depth camera array based on the imaging range of the depth camera array;

[0052] S1.3 arranges the depth camera array according to the set position. After the arrangement is completed, the depth camera array is moved in the chicken house to obtain several internal scene images at different angles in the chicken house.

[0053] Furthermore, the process of setting the position of the depth camera array based on the imaging range of the depth camera array in step S1.2 further includes the steps of:

[0054] S1.2.1 A camera rail array is provided around the chicken house on the top of the chicken house for horizontal movement of the depth camera array;

[0055] S1.2.2 Obtain the positions of the four corners and the center of the top of the chicken house, and determine the installation position of the camera slide array according to the positions of the four corners and the center of the top of the chicken house and the imaging range of the depth camera array. After the camera slide array is installed, install a lifting bracket on the camera slide array for vertical movement of the depth camera array, and finally set the depth camera array on the lifting bracket. In this embodiment, the depth camera array moves horizontally around the chicken house through the camera slide array and moves vertically through the lifting bracket, so as to realize multi-directional 360° panoramic shooting of the chickens in the chicken house, thereby realizing multi-directional three-dimensional detection of chicken heads.

[0056] It can be understood that this embodiment can quickly capture the development of the comb through multi-directional 360° panoramic detection of comb characteristics, thereby detecting diseases of chickens at an early stage and taking necessary preventive measures. By analyzing the size, color and shape data of the comb, the growth and nutritional status of the chickens can be understood, which helps farmers adjust feed formulas, stocking density and other management measures, optimize the breeding environment, and improve breeding efficiency. At the same time, through multi-directional 360° panoramic detection of the comb, data support can be provided for the genetic breeding and variety improvement of farmed chickens, helping farmers to select new varieties with excellent comb characteristics and achieve sustainable development of chicken farming.

[0057] As a preferred embodiment, in the process of using the depth camera array to obtain the internal scene image of the chicken house in step S1, the light environment detection in the chicken house is realized by using a current controller, and the process of the depth camera array obtaining the image is supplemented with light based on the detection result, so that the depth camera array can obtain an image with better clarity; the specific implementation process is as follows:

[0058] The light environment signal in the chicken house is received and converted to obtain a light environment control digital signal of the chicken house. Based on the light environment control digital signal, an infrared fill light device is used to fill light in the process of acquiring images by the depth camera array.

[0059] It is understandable that the depth camera array is provided with an infrared fill light device, and the acquired light environment control digital signal is connected to the control end of the infrared fill light device, and converted into an analog control signal suitable for the infrared fill light device through a digital-to-analog converter to ensure that the light environment control signal is compatible with the control interface of the infrared fill light device. By setting up an infrared fill light device to control the light environment and fill light for the depth camera array to acquire images, it is easy to obtain higher quality images and improve the stability and consistency of image acquisition.

[0060] Furthermore, the step of converting the light environment signal to obtain a light environment control digital signal for the chicken house includes:

[0061] The light environment signal is nonlinearly converted to obtain a light environment control electrical signal.

[0062] The light environment control electrical signal is converted into a light environment control digital signal through a data acquisition card.

[0063] Furthermore, the method for performing nonlinear conversion on the light environment signal is as follows:

[0064]

[0065] In the formula, A represents the amplitude of the signal, a(m) represents the symbol of the signal, p(t) represents the shaping function, and f c represents the carrier frequency of the signal, Represents the phase of the signal. After nonlinear transformation of the above formula, we get:

[0066]

[0067] The multipath space is constructed for the converted light environment signal as follows:

[0068]

[0069] In the formula, Q is the number of sampling points, K is the maximum delay, and the maximum detection distance R max / c to get, where x reci (t) is the reference signal, R max is the maximum detection distance, c is the speed of light;

[0070] Then, the least squares method is used to suppress the direct wave and its multipath to obtain the final regulated current signal. sur -X ref ·α|| 2 Convert to get:

[0071] Substitute α estim, and the final regulating current signal is obtained:

[0072]

[0073] Among them, S sur is the echo channel signal, α is the adaptive weight, α estim is the estimated value of α, X H ref For X ref The transpose of S other is the final regulated current signal in the echo channel.

[0074] It can be understood that this embodiment realizes the control of fill light in the process of acquiring images by the depth camera array through the above process, which is conducive to the depth camera array to obtain clearer images of the chicken house environment, and provide more accurate and reliable data support for the chicken farm, helping farmers to better understand the growth and egg-laying conditions of chickens, and ensure that the light intensity, light time and light quality in the chicken house meet the growth and egg-laying requirements of chickens, thereby improving the physiological health and production performance of chickens.

[0075] As a preferred embodiment, the process of performing color correction on the internal scene image based on a standard colorimetric card in step S2 comprises the following steps:

[0076] S2.1 constructing a mapping relationship matrix between the RGB values ​​of the standard colorimetric card and the RGB values ​​of the internal scene image by using a polynomial regression method;

[0077] S2.2 solves the mapping relationship matrix using the least square method;

[0078] S2.3 substitutes the RGB value of the internal scene image into the mapping relationship matrix, performs color correction based on the RGB value of the standard colorimetric card, and outputs the corrected RGB value of the internal scene image.

[0079] It can be understood that this embodiment performs color correction on the image after collecting the image of the internal scene of the chicken house, which can ensure that the color, skin condition, comb size and other characteristics of the chickens in the image are presented more realistically and accurately, helping veterinarians or breeders to more accurately judge the health status of the chickens, more accurately identify the number, location, behavior and other information of the chickens, so as to take timely treatment measures, and help to realize automated feeding, drinking water, cleaning, etc., and improve the automation level and efficiency of breeding.

[0080] As a preferred embodiment, the process of performing three-dimensional modeling of the chicken head in the chicken house based on the color-corrected internal scene image in step S3 includes:

[0081] S3.1 identifying the chicken head in the internal scene image based on an image segmentation model and an image detection model in a deep learning network, wherein the image segmentation model includes a U-Net model, and the image detection model mainly includes a YOLO model;

[0082] Furthermore, the YOLO model and the U-Net model are used to perform image segmentation on the internal scene image. The specific process includes: first, the internal scene image is normalized to meet the model input requirements, and then the processed internal scene image is divided into a training set and a verification set, and the chicken head in the training set is labeled. After the labeling is completed, the model parameters of the YOLO model are initialized, and the labeled training set is input into the YOLO model. A loss function is set to evaluate the difference between the prediction result of the model and the actual labeling, and a gradient descent algorithm is used to update the parameters of the YOLO model to minimize the loss function. After multiple iterative trainings, when the loss function tends to be minimized, the training process of the YOLO model is completed, and the output YOLO model parameters are used to detect the remaining internal scene images.

[0083] After the detection process of the chicken head is completed, the YOLO model completes the annotation of the chicken head position in each internal scene image. At this time, the segmentation process of the chicken head image is realized by the U-Net model. First, the U-Net model is trained. Since the present embodiment has realized the detection process of the chicken head by the YOLO model, it is only necessary to perform a simple model training process on the U-Net model (for example, only divide 5% of the data set as the training set). After the model training is completed, the internal scene image with the marked chicken head position is input into the model. The encoder part extracts the chicken head features in the image through the convolution layer and the pooling layer, and gradually reduces the size of the feature map, and restores the image details from the features extracted by the encoder through the decoder, and gradually increases the size of the feature map until it is the same as the original image size. Finally, on the basis of annotating the chicken head, the U-Net model outputs the probability that the pixel value in each image belongs to the chicken head feature. At this time, the image segmentation can be completed according to the probability value, and the segmentation result is visualized by OpenCV, so as to obtain an accurate chicken head feature image.

[0084] S3.2 aligns the chicken head images at different angles based on the point cloud registration algorithm, and performs three-dimensional reconstruction of the chicken head, thereby completing the construction process of the three-dimensional model of the chicken head.

[0085] Furthermore, the chicken head images acquired from different angles by the depth camera array are accurately fused, and an algorithm based on feature point matching and weighted averaging is used to reduce the error in the fusion process and construct a smoother and more accurate three-dimensional model of the chicken head image.

[0086] It can be understood that this embodiment combines the YOLO model and the U-Net model to complete the image segmentation, which can further improve the efficiency and accuracy of image annotation and segmentation. First, the YOLO model is used to detect the target of the image to obtain the position and category information of the target object, and then this information is used as the input of the U-Net model to perform pixel-level segmentation. This method can make full use of the advantages of the two models, annotate the target object in the image through the YOLO model, and provide high-quality annotation data for subsequent image segmentation, which helps the U-Net model to better learn the characteristics of the target object and improve the accuracy of segmentation;

[0087] In addition, compared with using the U-Net model alone for image segmentation, using the YOLO model for target detection first can greatly reduce the image area that needs to be processed, thereby reducing the computational cost, which is of great significance for processing large-scale data sets or real-time application scenarios.

[0088] As a preferred embodiment, the process of extracting comb features based on the three-dimensional model of the chicken head in step S4 includes:

[0089] S4.1 Smoothing and removing noise from the three-dimensional model of the chicken head;

[0090] S4.2 uses a Reeb graph to automatically segment the processed chicken head three-dimensional model. The Reeb graph shrinks each connected branch on the isosurface of the three-dimensional model into a point, thereby encoding the connectivity relationship between the critical points of different chicken organs, thereby constructing a topological structure graph. The model can be segmented through the topological structure graph to obtain a number of feature blocks;

[0091] S4.3 performs statistics on the feature blocks, and obtains the feature blocks belonging to the comb based on the statistical results, and finally uses the feature blocks belonging to the comb as the comb feature image.

[0092] As a preferred embodiment, the process of obtaining the comb measurement information based on the chicken head three-dimensional model and the comb feature image in step S5 includes:

[0093] S5.1 The shape of the comb feature block is used as the shape of the comb, and the color of the comb feature image is used as the color of the comb;

[0094] S5.2 Based on the internal scene image, obtain the reference object in the chicken house and obtain the actual size information of the reference object, perform feature extraction on the outline of the reference object based on OpenCV, obtain the pixel size of the reference object in the internal scene image, obtain the proportionality coefficient between the pixel size of the reference object and the actual size, and obtain the actual size of the comb based on the proportionality coefficient and the pixel value size of the comb feature image;

[0095] S5.3 constructs the comb measurement information based on the shape, color and actual size of the comb, thereby completing the comb detection process.

[0096] Furthermore, a large amount of image data of the normal and healthy state of the comb is collected to train the behavior pattern recognition model. By analyzing the shape, color and actual size sequence of the comb, the model can learn and recognize the image of the comb in a normal and healthy state. When information that does not match the normal image appears, the behavior pattern recognition model can issue an abnormal alarm in time and be ready to indicate the abnormal location. For example, when the shape, color and actual size of the comb change abnormally, the algorithm can quickly detect and alert the breeder to pay attention.

[0097] Furthermore, a deep learning algorithm is used to analyze the shape, color, actual size of the comb and the movement trajectory of the chicken's head to accurately estimate the chicken's head posture and then determine the chicken's behavioral state, such as whether it is foraging, drinking water, resting or exhibiting abnormal behavior.

[0098] Furthermore, based on the shape, color and actual size of the comb, a correlation model between physiological characteristics and diseases is established using machine learning algorithms. Through real-time monitoring and feature analysis of the shape, color and actual size of the comb, early diagnosis of diseases and health status assessment of chickens can be achieved. When its color gradually fades or turns red, combined with other physiological characteristics and behavioral data, it can be determined whether the chicken is infected with a disease or is in a sub-healthy state, and the probability of disease can be accurately determined.

[0099] It can be seen that real-time monitoring and analysis of the shape, color and actual size of the comb is helpful in confirming the status of the chickens during daily inspections and timely diagnosing the health status of the chickens.

[0100] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can still modify the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or perform equivalent replacements on some of the technical features thereof; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for measuring chicken head features based on 360-degree panoramic scanning, characterized in that: The following steps are involved: A depth camera array is used to obtain images of the internal scenes of the chicken house; Performing color correction on the internal scene image based on a standard colorimetric card; Perform three-dimensional modeling of the chicken head in the chicken house based on the color-corrected internal scene image to obtain a three-dimensional model of the chicken head; Extracting comb features based on the three-dimensional model of the chicken head to obtain a comb feature image; The comb measurement information is obtained based on the chicken head three-dimensional model and the comb characteristic image.

2. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: The process of using a depth camera array to acquire images of the interior scene of a chicken house includes: Acquire the imaging range of the depth camera array; Setting a depth camera array position based on the imaging range; The depth camera array is arranged according to the position, and internal scene images of several chicken houses at different angles are acquired based on the arranged depth camera array.

3. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 2, characterized in that: The process of setting the position of the depth camera array based on the imaging range includes: Set up a camera rail array around the top of the chicken house; Get the locations of the four corners and the center of the top of the chicken house; The installation position of the camera rail array is determined according to the positions of the four corners and the center of the top of the chicken house and the imaging range of the depth camera array. After the camera rail array is installed, a lifting bracket is installed on the camera rail array, and the depth camera array is set on the lifting bracket.

4. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: In the process of using several depth camera arrays to acquire images of the internal scenes of the chicken house, the light environment in the chicken house is detected and fill light is added to the process of the depth camera array acquiring images based on the detection results.

5. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 4, characterized in that: The process of detecting the light environment in the chicken house and supplementing the image acquired by the depth camera array based on the detection results includes: The light environment signal in the chicken house is received and converted to obtain a light environment control digital signal of the chicken house. Based on the light environment control digital signal, an infrared fill light device is used to fill light in the process of acquiring images by the depth camera array.

6. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 5, characterized in that: The steps of converting the light environment signal to obtain a light environment control digital signal for the chicken house include: Performing nonlinear conversion on the light environment signal to obtain a light environment control electrical signal; The light environment control electrical signal is converted into a light environment control digital signal through a data acquisition card.

7. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: The process of color correction of the internal scene image based on the standard colorimetric card includes: A mapping relationship matrix between the RGB values ​​of the standard colorimetric card and the RGB values ​​of the internal scene image is constructed by using polynomial regression. After the mapping relationship matrix is ​​solved by the least squares method, the RGB values ​​of the internal scene image are substituted into the mapping relationship matrix, color correction is performed based on the RGB values ​​of the standard colorimetric card, and the corrected RGB values ​​of the internal scene image are output.

8. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: The process of performing three-dimensional modeling of the chicken head in the chicken house based on the color-corrected internal scene image includes: The chicken head in the internal scene image is identified based on the image segmentation model and the image detection model in the deep learning network, and the point clouds of the chicken head at different angles are registered based on the point cloud registration algorithm, and the chicken head is three-dimensionally reconstructed to obtain the three-dimensional model of the chicken head.

9. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: The process of extracting the comb features based on the three-dimensional model of the chicken head includes: After smoothing and noise removal of the chicken head three-dimensional model, the processed chicken head three-dimensional model is automatically segmented using a Reeb graph to obtain a number of feature blocks, which are statistically analyzed, and feature blocks belonging to the comb are obtained based on the statistical results, and the comb feature image is obtained based on the feature blocks belonging to the comb.

10. The method for measuring chicken head features based on 360-degree panoramic scanning according to claim 1, characterized in that: The process of obtaining the comb measurement information based on the chicken head three-dimensional model and the comb feature image includes: The shape of the comb feature block is used as the shape of the comb, and the color of the comb feature image is used as the color of the comb; Based on the internal scene image, obtain a reference object in the chicken house and obtain the actual size information of the reference object, perform feature extraction on the outline of the reference object based on OpenCV, obtain the pixel size of the reference object in the internal scene image, obtain the ratio coefficient between the pixel size of the reference object and the actual size, and obtain the actual size of the comb based on the ratio coefficient and the pixel value size of the comb feature image; The comb measurement information is constructed based on the shape, color and actual size of the comb.

Images