Research On Key Technologies Of Intelligent Animal Experiment Box

Animal experiment has always been an important foundation in the field of science.The control of the environment,the monitoring of the experimental subject and the collection and processing of the experimental data are important parts of the experiment.The realization of these requirements mentioned above can not be separated from the assistance of the experimental equipment.With the continuous upgrading of experimental requirements,and in order to get accurate experimental data,the requirements of automation and intelligence of the experimental equipment are also increasing.At present,the research on animal experimental equipment at home and abroad is still insufficient,especially in the part of intelligent control and experimental subject monitoring.In this paper,an intelligent animal experimental box is designed and made based on the intelligent animal experimental equipment.Meanwhile,the key technologies of the box are deeply studied,including the intelligent control of the gas concentration inside the box and the real-time detection of the experimental subject.It has a substantial role in promoting the intelligent development of the experimental system.The main research work of this paper is as follows:(1)Based on the analysis of the characteristics of existing animal experiment equipment,an intelligent animal experiment box based on deep learning is designed.The combination of sensor and camera is used to control the internal variables of the experimental system and to collect the image data of the experimental body.Using improved GA-BP-PID algorithm to control internal environment variables.The YOLOv5 network model is improved to identify the experimental subject and record the data.According to the collected data,the behavior state of the experimental subject is analyzed and the results are returned to the researchers,so as to meet the requirements of automation and intelligence of the experimental box.(2)In order to meet the relative stability of the internal environment of the box,an improved GA-BP-PID control algorithm is proposed to realize the intelligent control of the internal environment variables.The algorithm trains BP neural network to realize the automatic optimization of PID controller parameters,and finds the optimal weight initial value of BP neural network through the improved genetic algorithm,so as to improve the convergence speed and stability of the network.(3)In this paper,the existing target detection algorithms are analyzed and studied in detail,and the experimental subject detection data set is made.The YOLOv5 network model is used as the monitoring model of the experimental subject to obtain the center point coordinates of the experimental body and the size information of the target frame.The structure of the model is simplified and the stemlock structure and OSA bottleneck structure are substituted into the original algorithm.Through the comparative analysis of the detection accuracy and real-time performance of the model,it is verified that the improved YOLOv5 model proposed in this paper has higher accuracy and faster speed.Pytorch machine learning library is used to verify the accuracy of the model,and the experimental results show that the detection accuracy of the improved yolov5 model has been improved greatly.(4)In this paper,the existing animal behavior state analysis is studied in detail,and combined with the experimental animal data obtained in the detection stage,the experimental animal behavior state analysis system is established.By using the position and size information of the object box,the data are processed and analyzed to get the movement and behavior information of the object.The behavior state is divided into three main analysis criteria: regional fitness,survival state and behavior analysis,which can help the experimenters master the survival state of the animal in real time,ensure the success rate of the experiment,and improve the intelligent degree of the experimental box.(5)MFC development architecture and OPENCV computer vision library are used to develop the software interface of the test platform to comprehensively test the detection accuracy and efficiency of the model in the actual experimental environment.Finally,the research work of this paper is summarized retrospectively,the shortcomings of this paper are pointed out,and the direction of follow-up research is briefly introduced.Although the key technology of intelligent animal experiment box is studied in this paper.for other application fields,the research results can also be used for reference,and can be further extended to robot and other related fields.