Research on real-time object detection based on Yolo algorithm

Abstract

Carrying controlled knives and guns and ammunition in public places is a serious threat to public safety, and the target proportion of images in monitoring is small, and the recognition background is complex. The existing monitoring methods in public places have problems that can not be well recognized automatically. Therefore, the YOLOV5 model is proposed to be applied in Intelligent Security to improve public safety. Focusing on the target detection of contraband in public under the application scenario of Intelligent Security, the characteristics and requirements of the detection task are analyzed, and the viewpoint of applying YOLOV5 algorithm to real-time detection of contraband is put forward. The history and basic principles of YOLO series algorithms and target detection are summarized. YOLOV5 performance test was carried out systematically to verify the feasibility of the method. Then, the detection dataset is made, format conversion and labeling are carried out, and the training data is expanded, and the related detection experiments are carried out on the computer. The results show that the contraband detection method based on YOLOV5 has great potential in practical application.