Laboratory Population Statistics And Management System Based On Deep Learning

At present,domestic general university laboratories generally have problems such as extensive personnel management and low scientific research efficiency.How to strengthen laboratory personnel management and improve scientific research efficiency are an area worth exploring.With the development of object detection technology and the widespread popularity of video surveillance,it is possible to strengthen the management of laboratory personnel through video surveillance systems and the object detection technology.Although the traditional object detection technology based on image processing has attain certain results,its detection accuracy is difficult to achieve an ideal state.Benefiting from the improvement in computer hardware performance,the object detection technology based on deep learning has made an important breakthrough,and can achieve extremely high detection accuracy while maintain the detection speed.'Therefore,how to achieve accurate image object detection in virtue of deep learning is becoming a hot research topic.By comparing and analyzing traditional algorithms and object detection techniques based on deep learning,this paper points out that the latter has better detection results and can be applied in more fields.Therefore,this paper closely combines the video surveillance of the laboratory with the object detection technology based on deep learning.Based on the analysis of the basic theory of deep learning,this paper focuses on the human head detection model in laboratory video surveillance.The main work includes:(1)The experimental environment was built.Including:A laboratory video surveillance system was established;A laboratory human head data set was built by collecting monitoring images in the laboratory;The software environment required for deep learning was configured.(2)The Faster R-CNN human head detection model based on different depth ResNet in the laboratory is implemented,which verifies that the feature extraction ability of the ResNet increases with the increase of network depth.Subsequently,the laboratory human head detection model based on Faster R-CNN,R-FCN and SSD models was implemented by ResNet152 respectively.The detection performance of the three in the laboratory was compared and analyzed.The R-FCN model based on ResNet152 was selected as the human head detection model in the laboratory.(3)A laboratory population statistics and management system have been designed and implemented.The system uses the laboratory head detection model to perform a detection on the laboratory personnel every minute,and uses the IoU filtering algorithm to correct the detection results,and finally locates the personnel and stores the relevant data.The system periodically send laboratory personnel statistics to the WeChat group.Users can also view real-time video surveillance and laboratory personnel statistics remotely within the campus network.