Photographer Behavior Recognition Under Static Background Video Monitoring

How to detect photographing behavior is a topic worth studying.In the existing methods,the detection of such behavior is lacking.Based on this situation,this paper carefully analyzes the photographing behavior,decomposes it into four main steps,and carries out the behavior definition.Through these four definitions,this paper builds a process framework for photographing behavior detection and involves the framework.The hand area extraction and manual area tracking algorithms have been improved.Firstly,this paper extracts the hand region from the skin color model by adding the secondary screening and motion detection of the skin color and the constraints of the human body's apparent features,which greatly suppresses the noise information and has better robustness.Secondly,the common target tracking algorithm,such as MeanShift,initializes the search point position of the candidate region in the current frame by using the location of the centroid point of the target region of the previous frame.This way,once the moving target is slightly too large,it is easy to lose the target,and the mo-vement of the hand is just faster.In this paper,a method for the initialization of candidate region search points is proposed.The vector of sparse optical flow is calculated by the framework of L-K pyramid,and the initialization of the search points of candidate regions is used to improve the tracking accuracy of the tracking algorithm.