Research on Target Recognition Based on Improved YOLOv3 Algorithm

With the rapid development of science and technology, intelligent technology has penetrated into all walks of life, and a full coverage monitoring system has been built to realize all-round video monitoring without dead ends. Video can be used for real-time monitoring, and once abnormal situations are found, they can be handled in time. At present, with the rapid development of artificial intelligence, machine vision has been widely used in all walks of life. In this paper, real-time detection is realized by obtaining monitoring video, which not only saves a lot of labor costs, but also can find out and take corresponding measures in time when there are abnormal behaviors such as falling or smoking in public places, thus reducing personal injuries and property losses in logistics warehousing. Based on the above situation, this paper takes the analysis of abnormal behavior in logistics warehouse as a foothold, selects YOLOv3 as the basic algorithm, and improves it, and puts forward YOLOv3-PS4. Because the detection of warehouse personnel in logistics warehouse is different from other multi-category detection, it is necessary to optimize the algorithm to achieve a refined effect, in order to improve the ability of feature extraction and positioning, and adapt to the detection task in logistics warehouse scene.