Non-contact poultry body size measuring method and system

Abstract

The invention discloses a non-contact poultry body size measuring method and system, and relates to the technical field of poultry breeding. According to the method, images are collected through a camera set when poultry enters a measurement area and the posture is stable, poultry body size data are obtained through the steps of image preprocessing, segmentation, feature recognition, three-dimensional reconstruction and the like, and meanwhile data storage and analysis including sex and month age estimation, health state evaluation and the like are conducted through a cloud server. And finally, outputting a detailed measurement report and a comprehensive statistical report. The measuring system is composed of a plurality of modules which work cooperatively. The system solves the problems that traditional manual measurement is low in efficiency and prone to causing stress to poultry and the prior art is insufficient in measurement precision and incomplete in data management and analysis, and has the beneficial effects of efficient and accurate measurement, reduction of poultry stress, comprehensive data analysis, convenient data management, visual use and the like; and a scientific and effective body size measurement solution is provided for poultry breeding.

Description

Technical Field

[0001] The invention belongs to the technical field of poultry breeding, and in particular relates to a non-contact poultry body size measurement method and system. Background Art

[0002] In the process of poultry breeding, accurate measurement of poultry body size is of great significance for understanding the growth status of poultry, optimizing the breeding environment, and breeding excellent varieties. Traditional poultry body size measurement mostly uses manual measurement methods, which is not only inefficient, but also easy to cause stress damage to poultry, affecting their growth and development. In recent years, with the development of technologies such as machine vision, non-contact poultry body size measurement technology has gradually emerged, but the existing technical solutions still have many problems, such as background interference, inaccurate measurement due to the changeable posture of poultry, and poor adaptability of different breeds. In addition, poultry likes to move and often gather together, and the interference between them poses a great challenge to the accurate measurement of body size. Therefore, it is particularly important to develop a non-contact, efficient and accurate poultry body size measurement method. Summary of the invention

[0003] The object of the present invention is to provide a non-contact poultry body size measurement method and system to solve the above-mentioned problems existing in the prior art.

[0004] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0005] The present invention is a non-contact method for measuring the body size of poultry, comprising the following steps:

[0006] S1 Image acquisition: Image acquisition: When the poultry enters the preset measurement area, the system automatically triggers the image acquisition command, uses the camera group to shoot the poultry image, cooperates with the image analysis algorithm to monitor the poultry movement status in real time, and locks the image for acquisition when the poultry posture is in a relatively stable state;

[0007] S2 Image preprocessing: Enhance contrast, remove noise and perform color correction on the collected digital images to improve image quality and facilitate subsequent analysis;

[0008] S3 Image segmentation and individual differentiation: Segment the preprocessed image to divide the individual areas of different birds, extract the shape features, color features, and texture features of each individual image, and construct a feature vector. Use a clustering algorithm to group images with similar features into one category, and establish an individual identification database to avoid repeated measurements through feature comparison.

[0009] S4 Feature Recognition: In the segmented individual images, the deep learning-based object detection algorithm is used to accurately locate the key body measurement points of the birds;

[0010] S5 3D reconstruction: Based on the identified key feature points, combined with multi-view image information or depth sensor data, a 3D model of the bird is constructed through a 3D reconstruction algorithm;

[0011] S6 Body size measurement: In the three-dimensional model, various body size data of poultry are obtained by calculation according to the preset body size parameters;

[0012] S7 Data Storage and Analysis: The measured body size data is stored in real time in the local database and uploaded to the cloud server simultaneously. The cloud server uses the intelligent analysis module to estimate the gender and age of the birds, count the population distribution, and evaluate the health status;

[0013] S8 Measurement result output: Generates a detailed measurement report including but not limited to various body size data, estimated gender, age, breed, and health status assessment results, and provides a comprehensive statistical report.

[0014] As a preferred technical solution of the present invention, in step S1, the camera group includes multiple high-definition cameras distributed above and on the sides of the poultry activity area, which can capture poultry images from different angles, and is equipped with a ring light source that can adaptively adjust the brightness according to the ambient light intensity.

[0015] As a preferred technical solution of the present invention, in step S2, the specific operations are as follows:

[0016] The background subtraction algorithm is used to combine the pre-collected and stored background image of the measurement area to calculate the foreground image to remove the interference of background debris;

[0017] Then the foreground image is grayed and converted into a grayscale image according to the grayscale value calculation formula;

[0018] Use Gaussian filtering algorithm to filter the grayscale image to enhance image contrast and remove noise.

[0019] As a preferred technical solution of the present invention, in step S3, image segmentation adopts a semantic segmentation algorithm based on deep learning, is trained through a cross entropy loss function, and uses a back propagation algorithm to adjust model parameters.

[0020] As a preferred technical solution of the present invention, in step S4, body size parameters are calculated based on the poultry geometric model and preset measurement rules, and the deep learning algorithm is trained using massive image data covering a variety of poultry breeds, growth stages and postures.

[0021] As a preferred technical solution of the present invention, in step S5, if multi-view image information is used for three-dimensional reconstruction, the triangulation principle is used; if depth sensor data is used, the three-dimensional model is constructed based on the spatial position information represented by different numerical values ​​in the depth image and combined with key feature points.

[0022] As a preferred technical solution of the present invention, in S7, the local database adopts a lightweight database management system. For farms with large amounts of data, the data is regularly backed up to the cloud for long-term storage, and individual data that no longer exists or is no longer needed in the database is regularly cleared.

[0023] As a preferred technical solution of the present invention, in step S7, when the cloud server estimates the gender and age of the poultry using the intelligent analysis module:

[0024] In terms of age estimation, a linear regression model between body size data and age was established;

[0025] In terms of gender determination, specific feature recognition algorithms are used;

[0026] When assessing health status, compare the measured body size data with the standard body size range of healthy poultry of the same breed and age, and compare it with the poultry's movement posture and feather gloss image features.

[0027] The present invention also provides a non-contact poultry body size measurement system, including an image acquisition module, an image preprocessing module, an image segmentation and individual differentiation module, a feature recognition module, a three-dimensional reconstruction module, a body size measurement module, a data storage and analysis module, and a measurement result output module. The modules work together to realize the above-mentioned non-contact poultry body size measurement method.

[0028] The present invention has the following beneficial effects:

[0029] Using multiple high-definition cameras to shoot from different angles, combined with image analysis algorithms to collect images when the poultry is stable, and with advanced image preprocessing, segmentation, feature recognition and 3D reconstruction algorithms, it is possible to accurately obtain poultry body size data. Compared with traditional manual measurement, the efficiency is greatly improved, the measurement accuracy is higher, and the error is reduced.

[0030] Reduce poultry stress: The non-contact measurement method avoids direct human contact with poultry, greatly reducing the stress impact on poultry, which is beneficial to the normal growth and development of poultry and ensures breeding benefits.

[0031] Comprehensive data analysis: The system can not only measure body size data, but also estimate the gender and age of poultry through the intelligent analysis module of the cloud server, count population distribution, and evaluate health status, providing farmers with comprehensive breeding data support, which helps to make scientific decisions and optimize breeding management.

[0032] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0034] Figure 1 It is a schematic diagram of the process mentioned in the present invention;

[0035] Figure 2 A schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] like Figure 1 A non-contact poultry body measurement method is shown, comprising the following steps:

[0038] S1 Image acquisition: Image acquisition: When the poultry enters the preset measurement area, the system automatically triggers the image acquisition command, uses the camera group to shoot the poultry image, cooperates with the image analysis algorithm to monitor the poultry movement status in real time, and locks the image for acquisition when the poultry posture is in a relatively stable state;

[0039] S2 Image preprocessing: Enhance contrast, remove noise and perform color correction on the collected digital images to improve image quality and facilitate subsequent analysis;

[0040] S3 Image segmentation and individual differentiation: Segment the preprocessed image to divide the individual areas of different birds, extract the shape features, color features, and texture features of each individual image, and construct a feature vector. Use a clustering algorithm to group images with similar features into one category, and establish an individual identification database to avoid repeated measurements through feature comparison.

[0041] S4 Feature Recognition: In the segmented individual images, the deep learning-based object detection algorithm is used to accurately locate the key body measurement points of the birds;

[0042] S5 3D reconstruction: Based on the identified key feature points, combined with multi-view image information or depth sensor data, a 3D model of the bird is constructed through a 3D reconstruction algorithm;

[0043] S6 Body size measurement: In the three-dimensional model, various body size data of poultry are obtained by calculation according to the preset body size parameters;

[0044] S7 Data Storage and Analysis: The measured body size data is stored in real time in the local database and uploaded to the cloud server simultaneously. The cloud server uses the intelligent analysis module to estimate the gender and age of the birds, count the population distribution, and evaluate the health status;

[0045] S8 Measurement result output: Generates a detailed measurement report including but not limited to various body size data, estimated gender, age, breed, and health status assessment results, and provides a comprehensive statistical report.

[0046] like Figure 2 As shown: The present invention also provides a non-contact poultry body size measurement system, including an image acquisition module, an image preprocessing module, an image segmentation and individual differentiation module, a feature recognition module, a three-dimensional reconstruction module, a body size measurement module, a data storage and analysis module, and a measurement result output module. The modules work together to realize the above-mentioned non-contact poultry body size measurement method.

[0047] A specific embodiment of the invention is as follows:

[0048] Specific implementation steps

[0049] (I) System installation and initialization

[0050] In a farm covering an area of ​​5,000 square meters, with 10 large chicken houses, each chicken house raises an average of 5,000 laying hens, a total of about 50,000 laying hens, and mainly raises two varieties of laying hens, Hy-Line Brown and Roman Pink. According to the layout of the chicken house, two high-definition cameras are installed in the center above the chicken house for overhead shooting, and three high-definition cameras are installed on the walls on both sides for side shooting. All cameras are equipped with high-speed shutters with a speed of more than 1 / 1000 seconds to ensure that the movement posture of laying hens can be clearly captured. A ring light source with a power of 500W and a brightness that can be automatically adjusted between 100-1000 lux is installed to provide stable and uniform lighting for image acquisition. A camera with a depth sensor is set between every two adjacent cameras to obtain depth information of laying hens. At the same time, a protective device with a protection level of IP65 is installed to protect the camera and light source from dust, water vapor, etc. in the breeding environment, and the camera and light source are fixed in a suitable position using an adjustable support structure to ensure that the activity area of ​​laying hens can be fully covered.

[0051] The high-performance GPU is used in combination with a dedicated image processing algorithm chip, which has a built-in deep learning algorithm optimized for poultry body measurement. At the same time, the intelligent analysis module is integrated and parallel computing technology is used to quickly process large amounts of image data and analyze measurement results.

[0052] When there is no one in the chicken house and no laying hens are active, 100 background images of different angles and lighting conditions are collected and entered into the system as basic data for the background subtraction algorithm. The expected body size indicators of Hy-Line Brown and Roman Pink laying hens at different growth stages are entered. For example, at 16 weeks of age, the expected height of Hy-Line Brown laying hens is 18-20 cm and the body length is 28-30 cm; the expected height of Roman Pink laying hens is 17-19 cm and the body length is 27-29 cm, providing a reference standard for subsequent data analysis and evaluation.

[0053] (II) Measurement process

[0054] Image acquisition: Image acquisition is performed at 9 a.m., 2 p.m., and 7 p.m. every day, and each acquisition lasts 30 minutes. When the laying hens enter the measurement area, 10 cameras simultaneously capture images of the laying hens from different angles, and the ring light source automatically lights up and adjusts to the appropriate brightness according to the ambient light intensity. The system monitors the movement status of the laying hens in real time. When the movement speed of the laying hens is detected to be lower than 5 cm / s, for example, when the laying hens are eating, drinking, or taking a short rest, the system can accurately capture images that meet the measurement requirements, lock the images for acquisition, and each camera acquires about 2,000 images each time. Image preprocessing: The acquired images are quickly transmitted to the image preprocessing module of the data processing unit. First, the background subtraction algorithm is used to generate the standard background image B(x, y) by the mean method in combination with the 100 background images collected in advance, and then the foreground image F(x, y) is calculated by the formula F(x, y) = I(x, y)-B(x, y) to remove the interference of background debris. Then, the foreground image is grayed according to the gray value calculation formula and converted into a grayscale image. Finally, the Gaussian filtering algorithm is used, the Gaussian kernel size is set to 5×5, the standard deviation is 1.5, and the grayscale image is filtered according to the formula to enhance the image contrast and remove noise, preparing for subsequent image analysis.

[0055] Image segmentation and individual differentiation: The preprocessed images enter the image segmentation and individual differentiation module and are segmented using the U-Net model based on deep learning. This model uses the cross entropy loss function

[0056]

[0057] The training was performed and the back propagation algorithm was used to adjust the model parameters. The number of training rounds was 50 and the learning rate was 0.001. In order to improve the segmentation accuracy, an attention mechanism module was added to the model. By calculating the attention weight of each pixel, the model paid more attention to the key feature areas in the image, especially the overlapping parts of the laying hens. For the mutually occluded or overlapping parts, the depth information obtained by the depth sensor was combined for analysis. By setting an appropriate depth threshold T d In this implementation, T d Set to 0.5 meters, when the depth image D(x, y) <T d At the same time, the images taken by multiple cameras with different viewing angles are used for information fusion. Based on the principle of triangulation, it is assumed that the optical centers of the two cameras are O1 and O2 respectively, and the coordinates of the corresponding points in the two images are (x i, y1), (x2, y2), the baseline distance between the two cameras is known to be b, which is 1 meter in actual installation, and the focal length of the camera is f, f is 25 mm, then the three-dimensional coordinates of the spatial point P can be calculated by the formula to further determine the exact position and posture of each laying hen, so as to more accurately segment the individual. After segmenting the individual areas of different laying hens, the shape, color, and texture features of each individual image are extracted to construct a feature vector, and the K-Means clustering algorithm is used to calculate the Euclidean distance d

[0058]

[0059] Images with similar features are grouped into one category. At the same time, an individual identification database is established, and the cosine similarity algorithm is used for feature comparison to avoid repeated measurements.

[0060] Feature extraction and body size measurement: In the segmented individual images, the images enter the feature recognition module, and the YOLOv5 model based on deep learning is used to accurately locate the key body size measurement points of laying hens, such as the top of the head, the tip of the beak, the base of the wings, the tip of the tail, the foot joints, etc. The model divides the input image into grids, each of which is responsible for predicting a bounding box and the category probability of the object in the bounding box. The bounding box parameters are optimized through the regression loss function, and the CI oU loss function is used (where a is set to 0.25). According to the located measurement points, combined with the geometric model of laying hens and the preset measurement rules, body size parameters such as body height, body length, wingspan, and chest width are automatically calculated. For example, by measuring the vertical distance d from the top of the head to the foot joint in the image v , combined with the previously calibrated ratio k between image pixels and actual length (set to 0.1, i.e. 1 pixel represents 0.1 cm), the formula H = k × d v Considered quite tall.

[0061] Data storage and analysis: The measured body size data is stored in real time in a local database using the SQLite lightweight database management system. At the same time, the data is uploaded to the Alibaba Cloud server. The cloud server uses the intelligent analysis module to estimate the gender and age of laying hens based on the measured data and the preset analysis model. In terms of age estimation, a linear regression model between body size data and age is established.

[0062]

[0063] Where M is the estimated age in months, X i is the body size data, a iand b are regression coefficients obtained through training with a large amount of sample data. In this implementation, they are obtained through training with the body size data of 1,000 laying hens of different ages (where a1=0.5 corresponds to the body height coefficient, a2=0.3 corresponds to the body length coefficient, and b=5) for estimation. In terms of gender determination, a specific feature recognition algorithm is used, such as calculating the comb area A and comparing it with the set threshold A. threshold (Hai-Line Brown Egg Chicken A threshold Set to 3 square centimeters, Roman Pink Laying Chicken A threshold The gender of laying hens is determined by comparing the body size of the laying hens with the standard body size range of healthy laying hens of the same breed and age.

[0064] [S min , S max ], combined with the movement posture of laying hens and the glossiness of feathers (such as analyzing the grayscale variance of feathers σ 2 gray , the threshold is set to 100) and other image features to comprehensively judge whether the laying hen is healthy.

[0065] Output of measurement results: The system generates a detailed measurement report for each laying hen, including various body size data, estimated gender, age, breed, and health status assessment results. At the same time, it provides a comprehensive statistical report for the entire laying hen population, classifies and counts by gender, age, breed, and other dimensions, and displays the number of laying hens in each category, average body size data, and the number and proportion of healthy laying hens and laying hens with possible health problems. The population distribution and health status distribution are intuitively presented through visual charts such as bar charts and pie charts, and can be viewed through mobile phone APP, making it convenient for farm staff to obtain data at any time.

[0066] (III) Application of measurement results

[0067] Management of individual laying hens: Farm staff can check the measurement reports of individual laying hens at any time through the mobile phone APP to understand their growth status. For example, in the measurement of the 12th week, the Hy-Line brown laying hen numbered 001 was measured to be 16 cm in height and 26 cm in length. The linear regression model estimated that the age was about 11.5 weeks, and the comb area was calculated to be 3.2 square centimeters, which was determined to be male (Hy-Line brown laying hen). Compared with the standard body size range, the body height is slightly lower than the standard lower limit, and the health status assessment shows that the feather grayscale variance is 110, which is slightly higher than the threshold. The staff can conduct further inspections and treatments on the laying hen in a timely manner, and take targeted breeding measures, such as adjusting feed nutrition, observing behavior, etc., to ensure its healthy growth.

[0068] Group laying hen management: Farm managers can fully understand the growth of the laying hen group based on the comprehensive statistical reports provided by the system. For example, in the statistics of the 16th week, the average height of the Hy-Line Brown laying hen group was 19 cm, the body length was 29 cm, and the proportion of healthy laying hens was 90%; the average height of the Roman Pink laying hen group was 18 cm, the body length was 28 cm, and the proportion of healthy laying hens was 88%. When it is found that the average body size of laying hens in a certain age group is growing slowly, such as the slow growth of the height of Roman Pink laying hens aged 14-16 weeks, the managers can adjust the feed formula of laying hens in this age group in time to increase the protein and calcium content to meet their growth needs. When it is found that a certain breed of laying hens has more health problems, such as the high proportion of Roman Pink laying hens, the breeding management and disease prevention measures for the laying hens of this breed can be strengthened, the disinfection times of the chicken coops can be increased, the ventilation system can be adjusted, and the breeding environment can be improved, thereby improving the health level and breeding efficiency of the entire laying hen group.

[0069] Implementation Effect

[0070] Through practical application in the laying hen farm, this non-contact poultry body measurement system has effectively improved the efficiency and accuracy of laying hen body measurement, provided comprehensive and scientific breeding data support for the farm, helped farm managers make more reasonable breeding decisions, and significantly improved the breeding efficiency of laying hens. Within half a year after the introduction of the system, the average egg production rate of laying hens increased by 8%, the feed conversion rate increased by 5%, and the breeding cost decreased by 10%.

[0071] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0072] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A non-contact poultry body measurement method, characterized in that: The following steps are involved: S1 Image acquisition: When a bird enters the preset measurement area, the system automatically triggers the image acquisition command, uses a camera group equipped with a depth sensor to capture images of the bird, and uses an image analysis algorithm to monitor the bird's movement status in real time. When the bird's posture is in a relatively stable state, the image is locked for acquisition; S2 Image preprocessing: Enhance contrast, remove noise and perform color correction on the collected digital images to improve image quality and facilitate subsequent analysis; S3 Image segmentation and individual differentiation: Segment the preprocessed image to divide the individual areas of different birds, extract the shape features, color features, and texture features of each individual image, and construct a feature vector. Use a clustering algorithm to group images with similar features into one category, and establish an individual identification database to avoid repeated measurements through feature comparison. S4 Feature Recognition: In the segmented individual images, the deep learning-based object detection algorithm is used to accurately locate the key body measurement points of the birds; S5 3D reconstruction: Based on the identified key feature points, combined with multi-view image information or depth sensor data, a 3D model of the bird is constructed through a 3D reconstruction algorithm; S6 Body size measurement: In the three-dimensional model, various body size data of poultry are obtained by calculation according to the preset body size parameters; S7 Data Storage and Analysis: The measured body size data is stored in real time in the local database and uploaded to the cloud server simultaneously. The cloud server uses the intelligent analysis module to estimate the gender and age of the birds, count the population distribution, and evaluate the health status; S8 Measurement result output: Generates a detailed measurement report including but not limited to various body size data, estimated gender, age, breed, and health status assessment results, and provides a comprehensive statistical report.

2. A non-contact poultry body measurement method according to claim 1, characterized in that: In step S1, the camera group includes multiple high-definition cameras distributed above and on the sides of the poultry activity area, which can capture poultry images from different angles, and is equipped with a ring light source, which can adaptively adjust the brightness according to the ambient light intensity.

3. The non-contact poultry body measurement method according to claim 1, characterized in that: In step S2, the specific operations are as follows: The background subtraction algorithm is used to combine the pre-collected and stored background image of the measurement area to calculate the foreground image to remove the interference of background debris; Then the foreground image is grayed and converted into a grayscale image according to the grayscale value calculation formula; Use Gaussian filtering algorithm to filter the grayscale image to enhance image contrast and remove noise.

4. The non-contact poultry body measurement method according to claim 1, characterized in that: In step S3, the image segmentation adopts a semantic segmentation algorithm based on deep learning, is trained through a cross entropy loss function, and uses a back propagation algorithm to adjust model parameters.

5. The non-contact poultry body measurement method according to claim 1, characterized in that: In step S4, body size parameters are calculated based on the poultry geometric model and preset measurement rules, and the deep learning algorithm is trained using massive image data covering a variety of poultry breeds, growth stages and postures.

6. The non-contact poultry body measurement method according to claim 1, characterized in that: In step S5, if multi-view image information is used for three-dimensional reconstruction, the triangulation principle is used; if depth sensor data is used, the three-dimensional model is constructed based on the spatial position information represented by different values ​​in the depth image and combined with key feature points.

7. The non-contact poultry body measurement method according to claim 1, characterized in that: In S7, the local database adopts a lightweight database management system. For farms with large amounts of data, the data is regularly backed up to the cloud for long-term storage, and individual data that no longer exists or is no longer needed in the database is regularly cleared.

8. A non-contact poultry body measurement method according to claim 7, characterized in that: In step S7, when the cloud server estimates the gender and age of the poultry using the intelligent analysis module: In terms of age estimation, a linear regression model between body size data and age was established; In terms of gender determination, specific feature recognition algorithms are used; When assessing health status, compare the measured body size data with the standard body size range of healthy poultry of the same breed and age, and compare it with the poultry's movement posture and feather gloss image features.

9. A non-contact poultry body measurement method system, characterized in that: The method comprises an image acquisition module, an image preprocessing module, an image segmentation and individual differentiation module, a feature recognition module, a three-dimensional reconstruction module, a body size measurement module, a data storage and analysis module, and a measurement result output module. The modules work together to realize the non-contact poultry body size measurement method described in any one of claims 1 to 8.

Images