A method for identifying dead birds in a standardised cage house

Abstract

The present application belongs to the technical field of dead chicken identification, and relates to a dead chicken identification method for a standardized cage chicken house. The method comprises the following steps: S0, time synchronization of binocular visible light cameras and thermal infrared cameras; S1, registration of binocular visible light cameras and thermal infrared cameras; S2, detection of the motion state of each chicken by using binocular visible light cameras and a YOLOX network, which is used as a first dead chicken judgment decision factor; S3, obtaining the average temperature of each chicken according to the position coordinates of each chicken in the previous time RGB image combined with the current time thermal infrared image, which is used as a second dead chicken judgment decision factor; and S4, judging the health state of the chicken by comprehensively considering the first dead chicken judgment decision factor and the second dead chicken judgment decision factor, and outputting the final result.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of dead chicken identification, and relates to a dead chicken identification method for a standardized cage chicken house. BACKGROUND

[0002] In the broiler breeding industry, there are still many problems to be solved. According to investigations, the average mortality rate of broilers in more than 50,000 large-scale broiler farms in China is about 10%, and most of the farms still rely on the eyes and experience of workers to judge the state of the chickens. This operation requires the workers to frequently patrol the chicken house, which not only causes a large amount of dust in the chicken house to be inhaled into the workers' bodies, causing harm to the workers, but also may cause omissions. Therefore, due to the poor working environment and high labor intensity, the labor cost is high in the poultry breeding industry. At the same time, the chicken disease has a group effect, and if the dead chickens in the chicken house cannot be identified and detected in time, it will spread rapidly in the chicken population, causing a large area of disease, which not only causes economic losses but also brings public health problems such as food safety. In order to solve the above problems, it is necessary to develop a high-precision and high-efficiency dead chicken identification method for a standardized cage chicken house.

[0003] At present, there are few studies on the specialized dead chicken identification method for a standardized cage chicken house in China, and it is still in the initial exploration stage. The Chinese utility model patent with the name of "cage laying hen health behavior monitoring system based on robot" (application number 201621224630.8) discloses a cage laying hen health behavior monitoring system based on robot. The system knocks the chicken cage to make the chickens stand up, and uses the method of counting the leg features of the chickens to judge whether the chickens are sick or dead. This method is not desirable. Knocking each chicken cage will disturb all the chickens, which is not conducive to the feeding and growth of the chickens. The Chinese invention patent with the name of "cage chicken intelligent inspection system and cruise control method of inspection robot thereof" (application number 201910350606.0) discloses an intelligent inspection system for cage chickens. The system uses a thermal imaging camera to collect the infrared thermograph of the entire chicken cage to judge the body temperature of the chickens. However, the judgment method uses the average temperature of the entire chicken cage to judge, and does not detect each chicken one by one, which may produce false positives and reduce efficiency, resulting in unnecessary manual labor. SUMMARY

[0004] In view of the problems of low accuracy of dead chicken identification in cage chicken houses and high labor cost, the purpose of the present application is to provide a dead chicken identification method for a standardized cage chicken house.

[0005] In order to achieve the above purpose, the present application provides the following technical solutions:

[0006] A method for identifying dead chickens in standardized caged chicken houses includes the following steps:

[0007] S0, time synchronization of binocular visible light camera and thermal infrared camera;

[0008] The Robot Operating System (ROS) is used to publish binocular visible light and thermal infrared images acquired by the binocular visible light camera and thermal infrared camera mounted on the standardized caged chicken house inspection robot. The binocular visible light camera publishes at a frequency of 10Hz, and the thermal infrared camera publishes at a frequency of 5Hz. At the same time, a time tag is added to the published images based on the clock of the Robot Operating System (ROS). According to the time tag, a topic blocking waiting method is used to synchronize the software timing. The current three images are considered as a synchronized image pair and are saved only when the time difference between any two binocular visible light images and thermal infrared images is less than 100ms.

[0009] S1, registration of binocular visible light camera and thermal infrared camera;

[0010] After the standardized caged chicken house inspection robot arrives at each inspection point on the inspection route, it registers the binocular visible light camera and thermal infrared camera to obtain the homography transformation matrix of the visible light image and the thermal infrared image.

[0011] S2. Use a binocular visible light camera and YOLOX network to detect the movement status of each chicken and use it as the first decision factor for judging the first dead chicken.

[0012] After camera registration is completed, the original RGB images of the chickens are acquired in real time using a binocular visible light camera. Using the existing trained YOLOX network, with the chicken head as the labeled target, the position coordinates of each chicken in the RGB image at the current time are obtained. The target features of the chicken head in the predicted bounding box of each chicken are used as templates for matching and tracking to obtain the position coordinates of each chicken in the RGB image at the next time. By subtracting the coordinates, the movement of each chicken is detected and used as the first decision factor for judging whether a chicken is dead. If a chicken does not move, it is judged to be in an abnormal state; otherwise, it is judged to be in a normal state.

[0013] S3. Based on the position coordinates of each chicken in the previous RGB image obtained in step S2 and the current thermal infrared image, obtain the average temperature of each chicken and use it as the second decision factor for judging dead chickens.

[0014] According to the position coordinates of each chicken in the current time RGB image obtained in step S2 and the homographic transformation matrix obtained in step S1, the position coordinates of each chicken in the thermal infrared image are calculated; according to the position coordinates of each chicken in the thermal infrared image, the average temperature of each chicken is calculated by using the infrared thermal image, and the average temperature is taken as the second dead chicken judgment decision factor; if the average temperature of a chicken is lower than 30 degrees or higher than 45 degrees, the chicken is judged as an abnormal state; otherwise, the chicken is judged as a normal state.

[0015] S4, the health status of the chicken is judged by comprehensively judging the first dead chicken judgment decision factor and the second dead chicken judgment decision factor, and the final result is output.

[0016] If the first dead chicken judgment decision factor and the second dead chicken judgment decision factor of the chicken are both abnormal states, the chicken is judged as a dead chicken; otherwise, the chicken is judged as a healthy chicken.

[0017] The step S1 specifically includes the following steps:

[0018] S1.1, a plurality of visible light images and thermal infrared images are shot, the camera is calibrated by Zhang Zhengyou camera calibration method, the intrinsic matrix M1 of the binocular left camera, the intrinsic matrix M2 of the binocular right camera, the intrinsic M of the thermal infrared camera thermal , the rotation matrix R between the binocular left and right cameras, the translation matrix T between the binocular left and right cameras, the rotation matrix R thermal of the thermal infrared camera relative to the binocular left camera and the translation matrix T thermmal of the thermal infrared camera relative to the binocular visible light camera are obtained;

[0019] S1.2, the ORB feature point detection method is used to extract feature points from the visible light image, and the coordinates of the feature points in the binocular left camera coordinate system Coordinates left and the coordinates of the feature points in the binocular right camera coordinate system Coordinates rig are obtained respectively;

[0020] S1.3, based on the relationship between the left and right cameras of the binocular camera and the world coordinate system (formula 1 and formula 2), the intrinsic matrix M1 of the binocular left camera, the intrinsic matrix M2 of the binocular right camera, the rotation matrix R between the binocular left and right cameras, the translation matrix T between the binocular left and right cameras obtained in step S1.1 are substituted into the feature points extracted in step S1.2, and the coordinates of each feature point in the world coordinate system Coordinates world are obtained by projection:

[0021]

[0022]

[0023] In Formula 1 and Formula 2, S1 and S2 are non-zero scale factors, I is a unit matrix, M1 is an intrinsic matrix of a left camera of the binocular; M2 is an intrinsic matrix of a right camera of the binocular; R is a rotation matrix between the left and right cameras of the binocular; and T is a translation matrix between the left and right cameras of the binocular;

[0024] S1.4, obtaining an intrinsic matrix M of the thermal infrared camera according to the step S1.1 thermal , a rotation matrix R of the thermal infrared camera relative to the left camera of the binocular thermal , a translation matrix T of the thermal infrared camera relative to the visible light camera of the binocular thermmal , performing back projection on the feature point coordinates Coordinates world in the world coordinate system obtained in the step S1.3 to obtain coordinates Coordinates thermal of the feature point in the thermal infrared coordinate system;

[0025]

[0026] In Formula 3, S is a non-zero scale factor;

[0027] S1.5, obtaining coordinates Coordinates left of the feature point in the left camera coordinate system of the binocular according to the step S1.2 and coordinates Coordinates thermal of the feature point in the thermal infrared coordinate system according to the step S1.4, and calculating a homographic transformation matrix H of the visible light image and the thermal infrared image based on Formula 4;

[0028]

[0029] In Formula 4, H is a homographic transformation matrix of the visible light image and the thermal infrared image; Coordinates the is coordinates of the feature point in the thermal infrared coordinate system; and Coordinates left is coordinates of the feature point in the left camera coordinate system of the binocular.

[0030] The standardization cage house inspection robot comprises a VLP-16 multi-line laser radar 1, an ultrasonic sensor 2, a robot chassis 3, an inertial measurement unit 4, an industrial computer 5, a lifting platform 6, a camera holder 7, and a binocular visible light+thermal infrared camera platform 8 with a built-in ROS.

[0031] The robot chassis 3 adopts a differential steering structure based on a two-wheel differential model; the VLP-16 multi-line laser radar 1, the inertial measurement unit 4 and the ultrasonic sensor 2 are respectively responsible for data acquisition of different sensor information, and send the sensor information to the industrial computer 5; the industrial computer 5 is responsible for robot autonomous positioning, mapping based on sensor information, and navigation decision formation and instruction issuing; the lifting platform 6 and the camera holder 7 are responsible for camera steering, so that the binocular visible light + thermal infrared camera platform 8 can shoot the picture of the whole chicken coop, and the binocular visible light + thermal infrared camera platform 8 is composed of a binocular visible light camera, a thermal infrared camera and an industrial computer with a built-in ROS system, and time synchronization is performed on the binocular visible light + thermal infrared camera.

[0032] Compared with the prior art, the beneficial effects of the present application are:

[0033] The present application performs time synchronization before camera registration, avoiding registration failure caused by different time synchronization; the position of the chicken in the RGB image is predicted through the existing YOLOX model; the motion state of the detected chicken is used as the first dead chicken judgment decision factor through front and rear frame comparison; through the registration of the binocular visible light camera and the thermal infrared camera, the projection of the chicken position from the binocular visible light image to the infrared thermal image is realized, and the average temperature of the recognized chicken area is calculated as the temperature of the chicken, which is used as the second dead chicken judgment decision factor; finally, if the detection results meet the first dead chicken judgment decision factor and the second dead chicken judgment decision factor at the same time, the chicken is judged to be a dead chicken. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 The flow chart of the dead chicken identification method for the standardized cage chicken house of the present application;

[0035] Figure 2 The existing YOLOX model structure;

[0036] Figure 3 The flow chart of the dead chicken identification method for the standardized cage chicken house of the present application;

[0037] Figure 4 The structure diagram of the inspection robot adopted in the embodiment of the present application;

[0038] Figure 5 The flow chart of the time synchronization method of the binocular camera and the infrared camera.

[0039] The reference signs in the drawings are:

[0040] 1 VLP-16 multi-line laser radar

[0041] 2 ultrasonic sensor

[0042] 3 robot chassis

[0043] 4inertial measurement unit (IMU)

[0044] 5industrial computer

[0045] 6elevating platform

[0046] 7camera holder

[0047] 8binocular visible light + thermal infrared camera platform DETAILED DESCRIPTION

[0048] The application will be further described below in conjunction with the accompanying drawings and examples.

[0049] As shown in Figure 1 and Figure 3 A method for identifying dead chickens in a standardized cage chicken house, comprising the following steps:

[0050] S0, time synchronization of binocular visible light cameras and thermal infrared cameras;

[0051] Taking thermal infrared images and visible light images during chicken movement will cause registration failure due to asynchronization of the two sensors; the images collected by different sensors will have a time difference, so time synchronization is needed before camera registration.

[0052] As shown in Figure 5 The application uses the Robot Operating System (ROS) to publish binocular visible light images and thermal infrared images obtained by binocular visible light cameras and thermal infrared cameras carried on the standardized cage chicken house inspection robot, wherein the binocular visible light cameras are published at a frequency of 10Hz, the thermal infrared cameras are published at a frequency of 5Hz, and a time label is added to the published images based on the clock of the Robot Operating System (ROS); according to the time label, software timing synchronization is performed in the topic blocking waiting mode, only when the time difference of any two images of the binocular visible light images and the thermal infrared images is less than 100ms, the current three images are considered as a synchronous image pair, and are saved;

[0053] S1, registration of binocular visible light cameras and thermal infrared cameras;

[0054] After the standardized cage chicken house inspection robot reaches each inspection point on the inspection route, the binocular visible light cameras and the thermal infrared cameras are registered, specifically including the following steps:

[0055] S1.1, multiple visible light images and thermal infrared images are taken, the cameras are calibrated by Zhang Zhengyou camera calibration method, and the intrinsic matrix M1 of the binocular left camera, the intrinsic matrix M2 of the binocular right camera, and the intrinsic M of the thermal infrared camera are obtained thermal, a rotation matrix R between the left and right cameras of the binocular, a translation matrix T between the left and right cameras of the binocular, a rotation matrix R of the thermal infrared camera relative to the left camera of the binocular therm and a translation matrix T of the thermal infrared camera relative to the binocular visible light camera thermmal ;

[0056] S1.2, using the ORB feature point detection method to extract feature points from the visible light image, respectively obtaining the coordinates of the feature points in the coordinate system of the left camera of the binocular Coordinates left and the coordinates of the feature points in the coordinate system of the right camera of the binocular Coordinates right ;

[0057] S1.3, based on the relationship between the left and right cameras of the binocular and the world coordinate system (formula 1 and formula 2), substituting the intrinsic matrix M1 of the left camera of the binocular, the intrinsic matrix M2 of the right camera of the binocular, the rotation matrix R between the left and right cameras of the binocular, the translation matrix T between the left and right cameras of the binocular obtained in step S1.1, and the feature points extracted in step S1.2, the coordinates of each feature point in the world coordinate system Coordinates world :

[0058]

[0059]

[0060] In formula 1 and formula 2, S1 and S2 are non-zero scale factors, I is a unit matrix, M1 is the intrinsic matrix of the left camera of the binocular; M2 is the intrinsic matrix of the right camera of the binocular; R is the rotation matrix between the left and right cameras of the binocular; T is the translation matrix between the left and right cameras of the binocular.

[0061] S1.4, according to the intrinsic matrix M thermal of the thermal infrared camera obtained in step S1.1, the rotation matrix R thermal of the thermal infrared camera relative to the left camera of the binocular, and the translation matrix T thermmal of the thermal infrared camera relative to the binocular visible light camera, the feature point coordinates in the world coordinate system Coordinates world obtained in step S1.3 are back-projected to obtain the coordinates of the feature points in the thermal infrared coordinate system Coordinates thermal ;

[0062]

[0063] In formula 3, S is a non-zero scale factor.

[0064] S1.5, the coordinates of the feature points obtained according to step S1.2 in the binocular left camera coordinate system Coordinates left and the coordinates of the feature points obtained in step S1.4 in the thermal infrared coordinate system Coordinates therm Based on formula 4, the homographic transformation matrix H of the visible light image and the thermal infrared image can be obtained; and then the corresponding relationship between any point on the visible light image and the thermal infrared image can be obtained through H.

[0065]

[0066] In formula 4, H is the homographic transformation matrix of the visible light image and the thermal infrared image; Coordinates thermal is the coordinates of the feature points in the thermal infrared coordinate system; Coordinates left is the coordinates of the feature points in the binocular left camera coordinate system;

[0067] S2, using the binocular visible light camera and the YOLOX network to detect the motion state of each chicken, as a first dead chicken judgment decision factor;

[0068] After camera registration, the original RGB image of the chicken is obtained in real time using the binocular visible light camera, and the YOLOX network which has been trained is used to obtain the position coordinates of each chicken in the current time RGB image by taking the chicken head as a labeling target. The target feature of the chicken head in each chicken prediction box is matched and tracked as a template to obtain the position coordinates of each chicken in the next time RGB image. By subtracting the coordinates, it is detected whether each chicken is moving, and it is used as the first dead chicken judgment decision factor. If a chicken does not move, it is judged that the chicken is in an abnormal state; otherwise, it is judged that the chicken is in a normal state.

[0069] The present application adopts the existing YOLOX network as shown in Figure 2 The network structure and training process of the YOLOX network are not described herein.

[0070] S3, according to the position coordinates of each chicken in the previous time RGB image obtained in step S2, the average temperature of each chicken is obtained in combination with the current time thermal infrared image, and is used as a second dead chicken judgment decision factor;

[0071] According to the position coordinates of each chicken in the current time RGB image obtained in step S2 and the homographic transformation matrix obtained in step S1, the position coordinates of each chicken in the thermal infrared image are calculated. According to the position coordinates of each chicken in the thermal infrared image, the average temperature of each chicken is calculated using the infrared thermal image, and is used as a second dead chicken judgment decision factor. If the average temperature of a chicken is lower than 30 degrees or higher than 45 degrees, the chicken is judged to be in an abnormal state; otherwise, the chicken is judged to be in a normal state.

[0072] S4. Combine the decision factors for judging the first and second dead chickens to determine the health status of the chickens and output the final result;

[0073] If both the first and second decision factors for determining whether a chicken is sick or dead are in an abnormal state, then the chicken is determined to be sick or dead; otherwise, the chicken is determined to be healthy.

[0074] like Figure 4 As shown, the standardized caged chicken house inspection robot of this invention includes a VLP-16 multi-line lidar 1, an ultrasonic sensor 2, a robot chassis 3, an inertial measurement unit (IMU) 4, an industrial control computer 5, a lifting platform 6, a camera gimbal 7, and a binocular visible light + thermal infrared camera platform with built-in ROS 8.

[0075] The robot chassis 3 adopts a differential steering structure based on a two-wheel differential speed model. The omnidirectional wheels support the robot body. Compared to a four-wheel slip structure, the standard two-wheel differential speed does not produce self-rotation drift, ensuring stable movement. The VLP-16 multi-line LiDAR 1, Inertial Measurement Unit (IMU) 4, and ultrasonic sensor 2 are responsible for data acquisition from different sensors and sending the sensor information to the industrial control computer 5. The industrial control computer 5 is responsible for autonomous robot localization and mapping based on the sensor information, as well as navigation decision-making and command issuance. The lifting platform 6 and camera gimbal 7 are responsible for camera steering, enabling the binocular visible light + thermal infrared camera platform 8 to capture images of the entire chicken coop. The binocular visible light + thermal infrared camera platform 8 consists of a binocular visible light camera, a thermal infrared camera, and an industrial control computer with a built-in ROS system, which synchronizes the time of the binocular visible light + thermal infrared cameras.

Claims

1. A method for identifying dead birds in a standardised cage house, characterised in that, The method comprises the following steps: S0, time synchronization of binocular visible light cameras and thermal infrared cameras; Publish the binocular visible light images and thermal infrared images obtained by the binocular visible light cameras and the thermal infrared cameras carried on the standardized cage chicken house inspection robot using the robot operating system ROS, wherein the binocular visible light cameras are published at a frequency of 10 Hz, the thermal infrared cameras are published at a frequency of 5 Hz, and a time label is added to the published images based on the clock of the robot operating system ROS; according to the time label, software timing synchronization is performed in a topic blocking waiting manner, and only when the time difference of any two of the binocular visible light images and the thermal infrared images is less than 100 ms, the current three images are a synchronous image pair, and are saved; S1, registration of binocular visible light cameras and thermal infrared cameras; After the standardized cage chicken house inspection robot reaches each inspection point on the inspection route, the binocular visible light cameras and the thermal infrared cameras are registered to obtain a homographic transformation matrix of the visible light images and the thermal infrared images; S2, detecting the motion state of each chicken using the binocular visible light cameras and the YOLOX network, and taking it as a first dead chicken judgment decision factor; After the camera registration is completed, the original RGB images of the chickens are obtained in real time using the binocular visible light cameras, the YOLOX network is used to obtain the position coordinates of each chicken in the current time RGB image by taking the chicken head as a labeling target, the target features of the chicken head in each chicken prediction box are matched and tracked as a template to obtain the position coordinates of each chicken in the next time RGB image; by subtracting the coordinates, it is detected whether each chicken is moving, and it is taken as a first dead chicken judgment decision factor; if a chicken does not move, it is judged to be in an abnormal state; otherwise, it is judged to be in a normal state; S3, obtaining the average temperature of each chicken according to the position coordinates of each chicken in the previous time RGB image obtained in step S2 and the current time thermal infrared image, and taking it as a second dead chicken judgment decision factor; According to the position coordinates of each chicken in the current time RGB image obtained in step S2 and the homographic transformation matrix obtained in step S1, the position coordinates of each chicken in the thermal infrared image are calculated; according to the position coordinates of each chicken in the thermal infrared image, the average temperature of each chicken is calculated using the infrared thermal image, and is taken as a second dead chicken judgment decision factor; if the average temperature of a chicken is lower than 30 degrees or higher than 45 degrees, the chicken is judged to be in an abnormal state; otherwise, the chicken is judged to be in a normal state; S4, judging the health status of the chicken by comprehensively considering the first dead chicken judgment decision factor and the second dead chicken judgment decision factor, and outputting the final result; If the first dead chicken judgment decision factor and the second dead chicken judgment decision factor of the chicken are both abnormal states, the chicken is judged to be a dead chicken; otherwise, the chicken is judged to be a healthy chicken.

2. The method of claim 1, wherein, The step S1 specifically comprises the following steps: S1.

1. Capture multiple visible light and thermal infrared images. Calibrate the camera using Zhang Zhengyou's camera calibration method to obtain the intrinsic parameter matrix M1 of the left binocular camera, the intrinsic parameter matrix M2 of the right binocular camera, and the intrinsic parameter M of the thermal infrared camera. thermal The rotation matrix R between the left and right binocular cameras, the translation matrix T between the left and right binocular cameras, and the rotation matrix R of the thermal infrared camera relative to the left binocular camera. therm The translation matrix T of the thermal infrared camera relative to the binocular visible light camera therm ; S1.2, feature point extraction is performed on the visible light image by using an ORB feature point detection method to obtain coordinates Coordinates of the feature points in a left camera coordinate system of the binocular camera left and coordinates Coordinates of the feature points in a right camera coordinate system of the binocular camera righ ; S1.3, based on the relationship between the left and right cameras of the binocular camera and the world coordinate system, that is, formula 1 and formula 2, substituting the intrinsic parameter matrix M1 of the binocular left camera, the intrinsic parameter matrix M2 of the binocular right camera, the rotation matrix R between the binocular left and right cameras, the translation matrix T between the binocular left and right cameras obtained in step S1.1, and the feature points extracted in step S1.2, projection is performed to obtain the coordinate point Coordinates of each feature point in the world coordinate system world : In formula 1 and formula 2, S1 and S2 are non-zero scale factors, I is a unit matrix, M1 is an intrinsic matrix of a left camera of the binocular camera; M2 is an intrinsic matrix of a right camera of the binocular camera; R is a rotation matrix between the left and right cameras of the binocular camera; and T is a translation matrix between the left and right cameras of the binocular camera; S1.4, the intrinsic matrix M of the thermal infrared camera acquired according to step S1.1 thermal , the rotation matrix R of the thermal infrared camera relative to the binocular left camera the , the translation matrix T of the thermal infrared camera relative to the binocular visible light camera thermal , the feature point coordinates Coordinates in the world coordinate system acquired in step S1.3 world , the coordinates Coordinates of the feature point in the thermal infrared coordinate system are obtained by back projection thermal ; In formula 3, S is a non-zero scale factor; S1.5, coordinates of the feature points obtained according to step S1.2 in the coordinate system of the left binocular camera Coordinates left and the coordinates of the feature points obtained according to step S1.4 in the coordinate system of the thermal infrared camera Coordinates thermal , and the homographic transformation matrix H of the visible light image and the thermal infrared image is obtained based on formula 4; In formula 4, H is a homographic transformation matrix of the visible light image and the thermal infrared image; Coordinates thermal Coordinates left Coordinates 3. A standardized cage house inspection robot for use in a method according to claim 1, characterized in that, The standardized cage chicken house inspection robot comprises a VLP-16 multi-line laser radar (1), an ultrasonic sensor (2), a robot chassis (3), an inertial measurement unit (4), an industrial computer (5), a lifting platform (6), a camera holder (7) and a binocular visible light+thermal infrared camera platform (8) with a built-in ROS system. The robot chassis (3) adopts a differential steering structure based on a two-wheel differential model; the VLP-16 multi-line laser radar (1), the inertial measurement unit (4) and the ultrasonic sensor (2) are respectively responsible for data acquisition of different sensor information and send the sensor information to the industrial computer (5); the industrial computer (5) is responsible for autonomous positioning and mapping of the robot based on the sensor information and navigation decision formation and instruction issuing; the lifting platform (6) and the camera holder (7) are responsible for camera steering, so that the binocular visible light+thermal infrared camera platform (8) can shoot the picture of the entire chicken cage; the binocular visible light+thermal infrared camera platform (8) is composed of a binocular visible light camera, a thermal infrared camera and an industrial computer with a built-in ROS system, and time synchronization is performed on the binocular visible light+thermal infrared camera.

Images