A chicken breeding inspection robot device, control system and method

Abstract

The present application relates to a kind of chicken inspection robot equipment, control system and method.The present inspection system and method are aimed at the practical application scene of the breeding of multiple chickens, automatically complete the routine inspection work of the health survival state of chicken, abnormality judgment, data acquisition and storage;Its technical features are: from the hardware point of view, the present system includes inspection robot and computing center.The inspection robot is mainly composed of motion chassis, lifting module and telescopic module, sensor module and control system etc.;From the control system and implementation method point of view, it can be suitable for multiple breeding scenes, multiple inspection modes, manual or automatic inspection for different working modes of chicken farm;The data collected and the data of computing center communicate and store bidirectionally;The present application is reasonably designed, can control the inspection process of multiple breeding scenes, can accurately detect and identify the state of chicken under complex background conditions, reduce manual inspection, and improve economic benefit.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of animal breeding, and particularly relates to a chicken breeding inspection robot device, a control system and a method. BACKGROUND

[0002] At present, the poultry breeding mode gradually develops towards the trend of intensification and large scale, and the feeding mode of the breeding farm is mostly scale free-range and the use of ladder type or stacked cage raising.

[0003] In scale free-range, the breeding area is large, mostly flat area in the forest or mountain. Artificial needs to constantly patrol a large range, which is time-consuming and labor-intensive.

[0004] In the cage raising mode, the cage layer can be as high as 5 layers or more, and the feeding density is greatly improved, which brings many inconveniences to artificial management. For example, in daily management, the chicken flocks in each layer and each cage need to be inspected to understand the dynamics of the chicken flocks. If the chicken cage is more than 3 layers, it is time-consuming and labor-intensive to inspect.

[0005] The use of inspection robots instead of management personnel can change the slow, repetitive and boring inspection method into automatic processing, so that the management personnel can focus on other matters of the breeding farm, reduce labor cost and time cost, and improve the efficiency of agricultural production. One of the key tasks of chicken breeding inspection operation is the inspection of dead and sick chickens. At present, most chicken farms still use manual inspection methods. Whether it is large-scale broiler production or individual breeders, the above-mentioned manual inspection method mostly uses traditional methods such as wooden stick knocking and visual inspection, which has the characteristics of strong subjectivity, high labor intensity and low production efficiency, and at the same time, it may cause missed inspection. The disease is highly contagious in high-density breeding, and the mortality rate is high. If the sick and dead chickens are not treated in time, they will spread, rot and autolyze, producing a large amount of bacteria and increasing the disease or death rate of other chickens.

[0006] From the application demand, there are two kinds of inspection robots that can be used for reference. One is the inspection robot in traditional industries such as power, manufacturing and water conservancy. Their robots are mostly applied in their respective industries, and neither the hardware design nor the inspection mode can meet the actual scene of broiler farms. The second is the inspection robot for livestock and poultry breeding that has been developed, which has the problem of lack of universality. Different cage raising facilities in each chicken farm, such as cage size, cage raising layer number and spacing, will lead to the fact that the inspection robot cannot be applied, and the ordinary single inspection method cannot adopt different inspection schemes according to the actual different breeding environment.

[0007] Since the above schemes are not targeted, the false positive rate is high. Therefore, how to realize a general poultry breeding device, control system and method suitable for different scenes has important practical significance. SUMMARY

[0008] The application aims to provide a chicken breeding inspection robot device, a control system and a method to solve the technical problem of high false alarm rate in the use of the original inspection device.

[0009] To solve the above technical problems, the specific technical solutions of the application are as follows:

[0010] A chicken breeding inspection robot device mainly comprises a motion chassis, a lifting module and a telescopic module, a sensor module and a control system.

[0011] The motion chassis realizes the collection and judgment position of the inspection robot according to the instructions or signals.

[0012] The lifting module and the telescopic module realize the automatic operation of the sensors installed thereon in the vertical and horizontal directions.

[0013] The sensor module realizes the positioning of the chickens and the collection of the state information of the chickens and the breeding environment.

[0014] The control system uniformly controls the above-mentioned sub-modules, collects and summarizes the data, stores and transmits the data, and interacts with the computing center.

[0015] Further, the lifting module and the telescopic module are composed of a guide rail screw module and a motor, the motor output shaft and the screw are connected through a shaft coupling, and the rotation of the motor shaft drives the rotation of the screw to move the lifting module and the telescopic module.

[0016] The sensor module comprises a target recognition platform, a state collection module, a gas analysis module and an excitation device.

[0017] The target recognition platform is used to capture and identify the target and feed back to the control system for positioning.

[0018] Further, the state collection module is installed on the lifting module and the telescopic module, and the state collection module comprises a visual camera collector and an infrared collector.

[0019] The state collection module collects the images of the chicken's body state, activity level, lying state, standing state and feeding state, and the infrared collector collects the infrared thermal imaging of the chicken.

[0020] Further, the excitation device and the state collection module are integrated on the lifting module and the telescopic module.

[0021] Further, the excitation device adopts sound excitation, light excitation or heat excitation.

[0022] Sound stimulation, using explosive sound, the sound of natural enemies;

[0023] Light stimulation, using direct light, stroboscopic light;

[0024] Heat stimulation, using infrared light short time irradiation, microwave short time irradiation.

[0025] Further, the gas analysis module is installed on the moving chassis, and the collection and analysis device of the putrescine gas is included.

[0026] Further, the control system is called through the preset operation mode information in the system and the mode selection of the administrator, selects the application scene mode, distributes instructions to each subsystem for inspection operation, transmits the information collected by the sensor to the network center, receives the judgment result in the calculation center, and further operates each sub-module according to the preset scene mode. The administrator can operate the robot through the control system.

[0027] Further, the calculation center receives the transmission information collected by the sensor module summarized by the control system, compares and calculates by using the model data comparison of the data and the historical experience database, and when the threshold is triggered, alarms the administrator and records in the database. The comparison and calculation adopts a mature YOLOV5 model calculation;

[0028] Using state inference model and / or thermal imaging contrast model, the images collected by the visual camera collector and the infrared collector are compared with the model data in the historical experience database, and the threshold values obtained by learning the model are compared according to the health, abnormality, disease or death, to finally determine whether the target exists, the target quantity and the position, and to record and store;

[0029] If judged as healthy, record in the healthy database;

[0030] If judged as abnormal, start the stimulation device to stimulate the abnormal chicken, collect the images and infrared images again after stimulating the abnormal chicken, and compare the data again with the model before stimulation, and make a two-way comparison and judgment; if judged as healthy, record in the healthy database, and record once abnormal; if judged as disease or death, record in the disease or death database, and feedback to the administrator for processing;

[0031] If judged as death, record in the death database, and feedback to the administrator for processing;

[0032] The received gas analysis module collects the concentration data of putrescine, calls the data obtained by the wind speed collection device in advance, measures the air flow state, calls the scene comparison curve in the historical experience database, compares and judges whether the concentration of putrescine meets the breeding requirements, and when a certain threshold is reached, it is determined that the chicken dies and / or the chicken's excrement or other organic matter affecting the chicken's living environment in the environment is putrefied more than required, and records and feeds back to the administrator for processing.

[0033] A control system and method of a patrol robot device, comprising:

[0034] Step 1, select the use scene, the administrator inputs the scene state data, if there is no data input, the control system calls the parameters in the preset scene library to preset the device running state parameters, and feeds back to the computing center at the same time, and selects the corresponding calculation comparison model;

[0035] Step 2, the control system sends instructions to the mechanical motion part, and the mechanical motion part moves to the initial position of the use scene according to the instructions;

[0036] Step 3, the sensor module starts to enter the preparation collection mode;

[0037] Step 4, each sensor starts to collect data according to the instructions sent by the control system, and transmits the data to the control system;

[0038] Step 5, the control system collects the data of each sensor, transmits the data to the computing center, and the computing center adopts the calculation model to perform comparison calculation, the calculation result is stored, and the judgment result is recorded according to the preset program, and the result is recorded and fed back to the administrator or the threshold trigger comparison operation is performed;

[0039] Step 6, compare the calculation result with the preset trigger threshold value, if the trigger threshold value is reached, further judgment instructions will be sent to the control system; wherein the triggered threshold value is determined by the trigger value through the learning model through the data collected in the breeding scene through statistics and model learning in the early stage;

[0040] Step 7, after the control system receives the further judgment instruction, the further judgment program is executed;

[0041] Step 8, the judgment program is performed, and the state data is collected again, steps 3 to 5 are run, until all judgment results meet the final end condition, and the further judgment program does not need to be implemented again, and the whole process of patrol work is completed.

[0042] Further, the specific comparison content of step 5 includes the environment state judgment step 5.1 and the chicken state judgment step 5.1;

[0043] Step 5.1, the calculation center will collect the concentration data of putrescine in the environment by the gas analysis module, or collect the concentration data of the abnormal putrescine in the surrounding environment, and call the data obtained by the wind speed collection device to measure the air flow state, call the comparison curve in the historical experience database, compare to determine whether the concentration of putrescine meets the breeding requirements or there is a situation of dead chickens; when a certain threshold is reached, it is determined that the death of the chicken and / or the decomposition of the chicken's feces or other organic matter affecting the living environment of the chicken in the environment exceeds the requirement, and the record is fed back to the administrator for processing;

[0044] Step 5.2, the calculation center identifies the image collected by the visual camera collector and the infrared collector by using the state reasoning model and / or the thermal imaging comparison model, compares the image with the model data in the historical experience database, compares the threshold values obtained by the model learning for health, abnormality, disease or death, finally determines whether the target exists, the target quantity and the position, and records and stores;

[0045] Step 7, the further judgment procedure described in the step, when the threshold is triggered, the calculation center sends a further judgment instruction to the control system, and the control system instructs the excitation device to start the excitation device; after the excitation, the state after the excitation is collected, and the data is returned and compared with the state before the excitation to calculate;

[0046] If it is judged to be abnormal, the excitation device is started to excite, and the abnormal chicken is excited again to collect images and infrared images and compare the data with the data before the excitation to make a two-way comparison and judgment; if it is judged to be healthy, the health database is recorded, and the abnormality is recorded; if it is judged to be sick or dead, it is recorded in the sick or dead database, and feedback is given to the administrator for processing.

[0047] The chicken breeding inspection robot device, control system and method of the application have the following advantages:

[0048] 1. Suitable for different breeding scenes, including but not limited to broiler, laying hen and ornamental chicken breeding inspection, and more widely applicable to scenes;

[0049] 2. The mobile fixed-point combination method is adopted, and the wide-area collection and fixed-point collection can be compared with the model to reduce the calculation processing amount and improve the comparison and judgment accuracy;

[0050] 3. By adopting the cooperation of the motion chassis and the lifting module and the telescopic module, the sampling position of the state acquisition module can be changed, the three-dimensional point position in the running area of the robot is realized, the whole large range acquisition is realized, the layer-by-layer inspection and the point-by-point inspection are realized. The image distortion caused by the shooting angle or the imaging light in the case of using fixed image acquisition or whole range acquisition is reduced, the calculation model load is increased, and the analysis and processing error rate caused by the distortion image processing is high.

[0051] 4. In the threshold judgment in the model, the abnormal state is compared and detected by using the excitation mechanism, the probability of misjudgment and wrong judgment is greatly reduced, the workload is reduced, and unnecessary loss is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0052] The application will be further described below in combination with the drawings:

[0053] Figure 1 It is a whole structure schematic diagram of the inspection robot of the application;

[0054] Figure 2 It is a whole structure schematic diagram of the system of the application;

[0055] Figure 3 It is a system function and data interaction schematic diagram of the application;

[0056] Figure 4 It is an inspection mode schematic diagram of the inspection robot of the application;

[0057] In the figure: 1. Motion chassis, 2. Lifting module and telescopic module, 3. Sensor module, 31. Target recognition platform, 32. State acquisition module, 33. Gas analysis module, 34. Excitation device; 4. Control system. DETAILED DESCRIPTION

[0058] In order to better understand the purpose, structure and function of the application, the application of a kind of chicken breeding inspection robot device, control system and method will be further described in detail below in combination with the drawings.

[0059] The inspection robot system of the application comprises an inspection robot and a computing center.

[0060] The inspection robot mainly comprises a motion chassis 1, a lifting module and a telescopic module 2, a sensor module 3 and a control system 4, etc.

[0061] The motion chassis 1 realizes that the inspection robot moves to the acquisition and judgment position according to the instruction or signal.

[0062] The lifting module and the telescopic module 2 realize the automatic operation of the sensors installed thereon in vertical and horizontal directions.

[0063] The sensor module 3 realizes the positioning of the chickens and the collection of the state information of the chickens and the breeding environment.

[0064] The control system 4 controls the above-mentioned sub-modules uniformly, collects and summarizes the data, stores and transmits the data, and interacts with the computing center.

[0065] The lifting module and the telescopic module 2 are mainly composed of a guide rail screw module and a motor. The motor output shaft is connected with the screw through a shaft coupling. The rotation of the motor shaft drives the rotation of the screw to move the lifting module and the telescopic module 2. The sensors installed thereon realize the automatic operation in vertical and horizontal directions.

[0066] The sensor module 3 includes a target recognition platform 31, a state collection module 32, a gas analysis module 33, and an excitation device 34.

[0067] The target recognition platform 31 is used to capture the recognized target and feed back to the control system 4 for positioning.

[0068] The state collection module 32 includes one or both of a visual camera collector and an infrared collector. The visual camera collector is used to collect images of the chicken's body state, activity level, lying state, standing state, and eating state. The infrared collector collects infrared thermal imaging of the chicken in each state. The state collection module 32 is installed on the lifting module and the telescopic module 2, which can better align and focus the imaging device, make the shooting more accurate, and the imaging clearer, further reduce the algorithm load, and improve the judgment accuracy.

[0069] The excitation device 34 is integrated with the state collection module 32 on the lifting module and the telescopic module 2. Because most of the broiler and egg chicken breeding methods adopt cage breeding, most of the time adopts lying state, and does not stand completely, it will be similar to the body state when it is lying or not eating, which may cause image misjudgment.

[0070] The excitation device 34 stimulates the chickens to produce vigilance. At this time, the chickens will change their state through the excitation machine. Then, the images after excitation are collected and compared with the images before excitation to judge the health state of the chickens.

[0071] The excitation device 34 can use sound excitation, light excitation, or heat excitation.

[0072] The sound excitation can use a certain decibel sound or explosion sound, enemy sound, etc.

[0073] The light excitation can use direct strong light or stroboscopic strong light.

[0074] Thermal excitation can be achieved by short-term infrared irradiation or short-term microwave irradiation.

[0075] The gas analysis module 33 is installed on the moving chassis 1 and includes a gas collection and analysis device, specifically a component for detecting putrescine-like substances. Its function is to collect the concentration of putrescine in the chicken coop or the surrounding environment where chickens are observed closely. To further increase accuracy, a wind speed collection device is installed on the moving chassis 1 to calculate the ventilation status.

[0076] The collection and analysis device collects ambient gas into a sample collector, dissolves the collected gas in a base liquid to form a putrescine solution to be tested, and the concentration of putrescine in ambient air can be indirectly determined by an electrochemical detection method.

[0077] Electrochemical detection methods applicable to rapid detection include electrochemiluminescence detection based on the enhancement effect of putrescine on the electrochemiluminescence of Ru(bpy)32+, and electrochemical biosensor detection based on the principle that the reduction peak current value generated by the working electrode for measuring putrescine in putrescine solution increases with the increase of putrescine concentration; both methods are applicable to the detection requirements of this invention.

[0078] Control system 4, through preset operating mode information and mode selection by the administrator, selects the application scenario mode and distributes instructions to each subsystem for inspection operations. Simultaneously, it transmits and retrieves information collected by sensors to the network center. Receiving judgment results from the computing center, it further operates each submodule according to the preset scenario mode. The administrator can issue commands to the robot through control system 4.

[0079] The computing center receives and aggregates information collected by sensor modules 3 from the control system 4. It performs comparative calculations by comparing this data with model data in a historical experience database. When a threshold is triggered, an alarm is triggered and recorded in the database. The comparative calculations utilize the mature YOLOv5 model.

[0080] Using state reasoning models and / or thermal imaging comparison models, images captured by visual cameras and infrared cameras are identified and compared with model data in historical experience databases. Based on the thresholds for health, abnormality, illness, or death obtained after model learning, the existence, quantity, and location of targets in each state are determined, and statistical records are made and stored.

[0081] It is identified as a health record health database.

[0082] If the chicken is determined to be abnormal, the excitation device 34 is started to excite the abnormal chicken. After the excitation, the image and infrared image are collected again, and the data before and after the excitation are compared again by the model to determine the abnormality in two directions. If the chicken is determined to be healthy, the health record database is recorded, and the abnormality is recorded. If the chicken is determined to be sick or dead, the sick or dead database is recorded, and the administrator is fed back for processing.

[0083] If the chicken is determined to be dead, the death database is recorded, and the administrator is fed back for processing.

[0084] The received gas analysis module 33 collects the concentration data of putrescine, calls the data obtained by the wind speed collection device in advance, measures the air flow state, calls the comparison curve in the historical experience database, and compares to determine whether the concentration of putrescine meets the breeding requirements. When a certain threshold is reached, it is determined that the chicken is dead and / or the chicken's feces or other organic matter affecting the chicken's living environment is decomposed beyond the requirements, and the administrator is recorded and fed back for processing.

[0085] In addition to this normal automatic inspection, the administrator can manually set and send instructions to the control system 4 through the computing center to perform manual inspection, pay more attention to the chicken rack with more abnormal chickens in the record, and obtain the judgment information. The running path of the robot can also be controlled to perform manual inspection, or the data obtained or the alarm is suspected, and manual instructions are sent to collect multiple or one state parameter through the sensor module 3, and manual confirmation is performed through model comparison.

[0086] All the collected data records in the database are used as historical database for model comparison, and the learning sample is increased to further improve the amount of basic data in the use scene, and the judgment accuracy in the use scene is improved through continuous learning and optimization.

[0087] Example of large-scale free-range breeding scene:

[0088] In the large-scale free-range breeding state, the free-range mode is selected first, and the whole system and each subsystem are preset according to the selected mode. At the same time, the corresponding calculation and comparison model is selected by feeding back to the computing center. First, the image collection device can be raised to capture a large range, and the state of the breeding chicken is determined as a whole. The computing center calculates the data collected by the image collection device, identifies the images collected by the visual camera collector and the infrared collector, compares the model data in the historical experience database, compares the threshold values obtained by the model learning for health, abnormality, sickness or death, and finally determines whether the target exists, the target quantity and position, and performs statistical recording and storage.

[0089] When the comparison of the state of the scene pre-design calculation model reaches the trigger threshold in the image returned by the control system 4, an abnormal instruction is sent to the robot, and the robot will move close to the abnormal chicken for observation to confirm the body posture, activity, and infrared imaging. At the same time, the action of the stimulating device 34 can be increased, and the stimulating device 34 is used to stimulate the farmed chicken to produce vigilance. At this time, the state of the chicken will be changed through the stimulation device. Then, the image after the stimulation is collected and compared with the image before the stimulation to judge the health status of the chicken.

[0090] If it is judged to be healthy, the health record database is recorded.

[0091] If it is judged to be abnormal, the stimulating device 34 is started to stimulate, and the image and infrared imaging after the stimulation of the abnormal chicken are collected again and compared with the data before the stimulation to make a two-way comparison and judgment. If it is judged to be healthy, the health record database is recorded, and it is once abnormal. If it is judged to be sick or dead, it is recorded in the sick or dead database and fed back to the administrator for processing.

[0092] If it is judged to be dead, it is recorded in the death database and fed back to the administrator for processing.

[0093] During the inspection process, the gas analysis module 33 collects the concentration data of putrescine in the environment, or collects the concentration data of putrescine in the surrounding area of the abnormal chicken during the close observation. At the same time, the data obtained by the wind speed collection device is called to measure the air flow state, and the comparison curve in the historical experience database is called to compare and determine whether the concentration of putrescine meets the breeding requirements or there is a situation of sick or dead chickens. When a certain threshold is reached, it is determined that the death of the chicken and / or the decomposition of the chicken manure or other organic matter in the environment affecting the living environment of the chicken exceeds the requirement, and the record is fed back to the administrator for processing.

[0094] Statistical data comparison of actual application scenarios:

[0095] Cage breeding scene example:

[0096] In the state of cage breeding, in order to improve the breeding density, multiple rows and multiple layers, and ladder type cage breeding chicken farm environment are often used. In such an environment, fixed image collection or overall range collection is used, and then image partition marking or imaging processing is used for analysis. It is easy to cause image distortion due to shooting angle or imaging light, resulting in high misjudgment rate of analysis and processing.

[0097] First, the cage breeding mode is selected, and the whole machine system and each subsystem are pre-set according to the selected mode, and feedback is given to the calculation center, and the corresponding cage breeding calculation comparison model is selected.

[0098] The system first calls the cage running mode, the management personnel can input the cage height, cage length and box layer, or automatically judge the cage height, cage length and layer through the target recognition platform 31, through the state acquisition module 32 equipped with lifting, the posture, activity, infrared image of each chicken can be collected layer by layer from low to high or from high to low, because the camera is designed to face each chicken cage, the image acquisition is carried out. The calculation center compares the image collected by the image acquisition device with the model data in the historical experience database by using state inference model and / or thermal imaging contrast model, compares the threshold value obtained by model learning for health, abnormality, disease or death, finally judges whether the target exists, the target quantity and position, and records and stores.

[0099] When the comparison between the image returned by the control system 4 and the scene pre-design calculation model reaches the trigger threshold, the robot is instructed to start the excitation device 34, the excitation device 34 stimulates the farmed chicken to produce vigilance, which will change the state of the chicken through excitation. Then compare the image after excitation with the image before excitation to judge the health status of the chicken.

[0100] If it is judged to be healthy, record the health database.

[0101] If it is judged to be abnormal, start the excitation device 34 to excite, and then collect the image and infrared image after excitation and compare them with the data before excitation to make a two-way comparison and judgment.

[0102] If it is judged to be sick or dead, record it in the sick or dead database and feedback to the administrator for processing.

[0103] If it is judged to be dead, record it in the death database and feedback to the administrator for processing.

[0104] Such a collection method reduces the load of later image processing and improves the imaging quality. Because the imaging angle of each image is consistent, the calculation of the comparison learning model will be more targeted, and the accuracy of the judgment will be greatly improved.

[0105] During the inspection process, the gas analysis module 33 collects the concentration data of putrescine in the environment, and at the same time, the data obtained by the wind speed collection device is called to measure the air flow state, and the comparison curve in the historical experience database is called to compare and judge whether the concentration of putrescine meets the breeding requirements or there is a situation of dead chickens. When a certain threshold is reached, it is judged that the death of the chicken and / or the decomposition of the chicken's excrement or other organic matter in the environment affecting the living environment of the chicken exceeds the requirement, and the record is fed back to the administrator for processing.

[0106] Further in the breeding of laying hens, the laying rate can be increased as a reference value. When the laying rate of each hen or the entire cage exceeds the average variance, the frequency of inspection of the hen or the entire cage can be increased to detect abnormal conditions in a timely manner. Or do cleaning and disposal or culling.

[0107] Data comparison of actual application scenarios:

[0108] From the above data comparison, it can be seen that the accuracy of the system is significantly improved compared to the previous fixed wide-angle collection and mobile wide-angle collection system. The false positive rate and false negative rate are greatly improved. The mortality rate is also significantly reduced. The analysis of the reasons should be due to the significant reduction in the false negative rate, which effectively controls the infection and impact of sick chickens on healthy chickens.

[0109] The historical experience database and usage scenario database mentioned in this paper, as well as the state comparison model and the reasoning model, are all various states and environmental data of breeding chickens accumulated in the previous breeding farm tracking research. After sorting, classification, labeling, summarizing, analyzing, model building, and model learning, the experience and scenario database and model library are finally established. The learning model used is the mature YOLOv5 network structure.

[0110] The specific method is to collect various state images of chickens; label the collected data set, including chicken position coordinate information, chicken size length and width information, and chicken category information and various state corresponding labels. Select the YOLOv5 network structure as the reference network. The Backbone of the basic network mainly includes the Focus structure to divide the data into training and validation sets and test sets, and the training and validation sets are divided into training sets and validation sets, which can complete all preprocessing before image input. Set the model training parameters, train and optimize the target detection model: build a virtual environment on the GPU server for training the model, after completion, input the training set into the YOLOv5 network structure for target detection model training, and after training is completed, the inference model for chicken state recognition detection is obtained. According to the effect obtained by verification, the model is optimized, and finally the model with the relatively best effect is obtained.

[0111] It is to be understood that the present application is described by way of example only, and that modifications or alterations can be made to the features and embodiments described without departing from the spirit and scope of the application. In addition, modifications can be made to the features and embodiments described to accommodate specific situations and materials without departing from the spirit and scope of the application. Accordingly, the application is not limited to the specific embodiments disclosed herein, but rather, the scope of the application includes all embodiments falling within the scope of the claims.

Claims

1. A chicken farming inspection robot device, characterized by, The utility model relates to a kind of poultry health monitoring system, including motion chassis (1), lifting module and telescopic module (2), sensor module (3) and control system (4). The motion chassis (1) realizes that poultry health monitoring system moves to collection, judging position according to instruction or signal. The lifting module and telescopic module (2) realize the automatic operation of the sensor vertically and horizontally installed on it, so that the sensor module (3) can be close to the collection in the three-dimensional space breeding position. The sensor module (3) realizes the positioning of chicken, the state information collection of chicken and breeding environment. The sensor module (3) includes target recognition platform (31), state collection module (32), gas analysis module (33) and excitation device (34). The excitation device (34) uses sound excitation, light excitation or thermal excitation. The control system (4) controls the motion chassis (1), lifting module and telescopic module (2) and sensor module (3) uniformly, and collects and summarizes the data, carries out data storage and transmission, and interacts with the computing center. The computing center receives the transmission information collected by the sensor module (3) summarized by the control system (4), compares the model data of the data and historical experience database, and calculates when the threshold is triggered. Using the mature YOLOV5 model to calculate, the state inference model and / or thermal imaging comparison model are used to identify the images collected by visual camera collector and infrared collector, compare with the model data in historical experience database, compare according to the threshold value obtained by model learning for health, abnormal, disease or death, finally determine whether the target exists, target quantity and position, and record statistics, store. If it is judged to be healthy, it is recorded to the healthy database. If it is judged to be abnormal, start the excitation device (34) to excite, collect images and infrared images again after excitation, and compare the data before and after excitation again, and judge by two-way comparison. If it is judged to be healthy, it is recorded to the healthy database. If it is judged to be disease or death, it is recorded to the disease or death database, and feedback to the administrator for processing. If it is judged to be death, it is recorded to the death database, and feedback to the administrator for processing. Receive the concentration data of putrescine collected by the gas analysis module (33), call the data obtained by the wind speed collection device in advance, measure the air flow state, call the scene comparison curve in the historical experience database for comparison to determine whether the concentration of putrescine meets the breeding requirements, and when a certain threshold is reached, determine that the chicken has died and / or the chicken's feces or other organic matter affecting the chicken's living environment has exceeded the requirements, and record and feedback to the administrator for processing.

2. The chicken breeding inspection robot device according to claim 1, wherein The lifting module and the telescopic module (2) are composed of a guide rail screw module and a motor, the motor output shaft is connected with the screw through a shaft coupling, and the motor shaft rotates to drive the screw to move the lifting module and the telescopic module (2); The target recognition platform (31) is used to capture and identify the target and feed back to the control system (4) for positioning.

3. The chicken breeding inspection robot device according to claim 2, wherein The state collection module (32) is installed on the lifting module and the telescopic module (2), and the state collection module (32) includes a visual camera collector and an infrared collector, the visual camera collector collects images of the chicken's body state, activity level, lying state, standing state, and feeding state, and the infrared collector collects infrared thermal imaging of the chicken, The control system (4) accepts manual instructions to control the operation of each module.

4. The chicken breeding inspection robot device according to claim 2, wherein The excitation device (34) and the state collection module (32) are integrated on the lifting module and the telescopic module (2), the excitation device (34) stimulates the breeding chicken to produce vigilance, and the state collection module (32) transmits the images collected after excitation back to the calculation center for comparison with the images before excitation to determine the health status of the chicken.

5. The chicken breeding inspection robot device according to claim 4, wherein Sound excitation uses explosive sound and the sound of natural enemies; Light excitation uses direct strong light and stroboscopic strong light; Heat excitation uses infrared light emission for short-time irradiation and microwave short-time irradiation.

6. The chicken breeding inspection robot device according to claim 2, wherein The gas analysis module (33) is installed on the motion chassis (1) and includes a putrescine gas collection and analysis device.

7. The chicken breeding inspection robot device according to any one of claims 1-6, wherein The control system (4) is called through the pre-set operation mode information in the system and the mode selection of the administrator, selects the application scene mode, distributes instructions to each subsystem for inspection operation, simultaneously transmits the information collected by the sensor to the network center for data transmission and calling, receives the judgment result in the calculation center, further operates each sub-module according to the pre-set scene mode, and the administrator can operate the robot through the control system (4).

8. A control method for the chicken breeding inspection robot device according to claim 1, wherein Step 1, select the use scene, the administrator inputs the scene state data, if there is no data input, the control system (4) calls the parameters in the preset scene library to preset the equipment running state parameters, and feeds back to the computing center, and selects the corresponding computing comparison model; Step 2, the control system (4) sends instructions to the mechanical movement part, and the mechanical movement part moves to the initial position of the use scene according to the instructions; Step 3, the sensor module (3) starts to enter the preparation collection mode; Step 4, each sensor starts to collect data according to the instructions sent by the control system (4), and transmits the data to the control system (4); Step 5, the control system (4) collects the data of each sensor, transmits it to the computing center, and uses the computing model to compare and calculate, stores the calculation result, and judges according to the preset program, records the result and feeds back to the administrator or performs the threshold trigger comparison operation; Step 6, compare the calculation result with the preset trigger threshold value, if the trigger threshold value is reached, send further judgment instructions to the control system (4); wherein the triggered threshold value is a trigger value determined by learning model through statistics and model learning of the data collected in the early stage in the breeding scene; Step 7, after the control system (4) receives the further judgment instruction, the further judgment program is executed; the further judgment program includes: the instruction excitation device excites the chicken of the trigger threshold value; after excitation, the running state after excitation is collected, and the data is returned to compare and calculate with the state before excitation, so as to perform bidirectional comparison and judgment; the excitation device uses sound excitation, light excitation or heat excitation; Step 8, the judgment program is performed, the state data is collected again, steps 3 to 5 are executed, until all judgment results meet the final end condition, and the further judgment program does not need to be executed again, and the whole process of one-time inspection is completed.

9. The control method according to claim 8, wherein the specific comparison content of step 5 includes environment state judgment step 5.1 and chicken state judgment step 5.1; Step 5.1, the computing center collects the concentration data of putrescine in the environment collected by the gas analysis module (33), or collects the concentration data of putrescine near the abnormal chicken, and calls the data obtained by the wind speed collection device to measure the air flow state, calls the scene comparison curve in the historical experience database, compares and judges whether the concentration of putrescine meets the breeding requirements or there is a situation of dead chicken; when a certain threshold value is reached, it is judged that the chicken is dead and / or the chicken manure or other organic matter affecting the living environment of the chicken in the environment is decomposed more than required, and the record is fed back to the administrator for processing; ​ Step 5.2, the calculation center identifies the image collected by the visual camera collector and the infrared collector by calculating the image collected by the image collection device, compares the model data in the historical experience database, compares the threshold values obtained after learning the model for health, abnormality, disease or death, and finally determines whether the target exists, the target quantity and the position, and records and stores the statistics; The further judgment program in step 7, when the threshold is triggered, the calculation center sends a further judgment instruction to the control system (4), the control system (4) instructs the excitation device (34) to excite the chicken that triggers the threshold; after excitation, the post-excitation state is collected, and the data is returned for comparison and calculation with the pre-excitation state; If judged to be healthy, record to the health database; If judged to be abnormal, start the excitation device (34) to excite, collect the image and infrared image of the abnormal chicken after excitation, and compare the data with the pre-excitation model again for bidirectional comparison and judgment; if judged to be healthy, record to the health database and record the abnormality; If judged to be sick or dead, record to the sick or dead database and feedback to the administrator for processing; If judged to be dead, record to the death database and feedback to the administrator for processing.

Images