Algorithm of Detection and Tracking for Laboratory Unsafe Behavior Based on Computer Vision

Computer vision gives computers the ability of visual perception similar to people, so that computers can feel the environment in the field of vision, understand the content of the feeling, and take corresponding actions. Object tracking is a problem of creating corresponding matching between successive image frames based on position, speed, shape, texture and color. Many security problems in laboratories are caused by unsafe behaviors, and most laboratories have built a complete video surveillance system based on computer vision, using these video surveillance data to identify unsafe behaviors. In this paper, image denoising based on morphological filter algorithm, moving target detection based on optical flow algorithm and moving target tracking based on Mean-Shift algorithm are used to provide key technical solutions for the study of unsafe behaviors in laboratories. The results show that the optical flow algorithm has good adaptability and is suitable for both static and dynamic background. If a more accurate similarity measurement function is used, the tracking time is shorter and the speed is faster.