Detection Method Research And System Development Of Laying Hen Behavior Based On Deep Learning

At present,although some laying broiler breeding equipment and environmental control technology have been rapidly developed,the health and welfare status of laying hens in the breeding process is still inefficient,labor intensive,and not conducive to healthy manual operations.Way to proceed.This approach still restricts the development of the laying hen industry to intensification and scale.Therefore,the research on the behavioral testing of laying hens is of great significance in the automated breeding and information farming of laying hens.However,due to the similarity of individual hens and the complicated breeding environment,few people use image technology to detect laying hens.This topic takes the laying hen as the research object,and conducts in-depth research on the behavior detection and key technologies of cage laying hens.The main work done is as follows:(1)Construct a function test system for laying hens.The research and design of the system's detection hardware platform was carried out,and the software system of the hens behavior detection system was designed.Complete the Pascal_VOC dataset production.According to the requirements,the image data is added with noise,translation,rotation and other operations for data amplification,and a reasonable and accurate training data set and test data set are prepared for the behavior test of the laying hen.(2)The principle and detection process of the current popular deep learning-based detection algorithms,such as Faster R-CNN,SSD,and YOLOV3,are studied.The comparison experiments of Faster R-CNN,SSD and YOLOV3 detection algorithms are carried out.The experimental results show that YOLOV3 is superior to the other two algorithms in detection speed and accuracy.However,in the actual detection based on the YOLOV3 algorithm,there may be a phenomenon such as stagnation and delay.(3)Based on the YOLOV3 detection algorithm,a lightweight TD-YOLOV3 detection algorithm is designed.Firstly,based on the YOLOV3 network structure,the lightweight T-YOLOV3 network structure is obtained by compression.Secondly,the RES module after Conv9 in the T-YOLOV3 network is replaced by the Dense block.After the Conv5,Conv7,Conv9,Conv12 conventional layer add the MLP structure.Finally,the K-means algorithm based cluster dimension optimization and training strategy optimization are used to train and test the data set.Experiments show that the TD-YOLOV3 algorithm not only meets the requirements in terms of accuracy,but also significantly improves the detection speed.(4)According to the research results of the behavior test of laying hens,an online detection system for laying hen behavior was designed based on TD-YOLOV3 algorithm and DJango front-end development technology.It has realized functional modules such as hens behavior detection and database.