Video Based Human Head and Shoulders Detection Using Machine Learning

With the development of artificial intelligence, the application of computer technology for analyzing the objects in images and videos has been more and more widely used. This paper describes the human head and shoulders based on Haar-like features, and realizes the detection of the head and shoulders in the video using machine learning. Firstly, this paper selects the ViBe background model method to extract the moving area in the video. Then, Haar-like features are selected to describe the human head and shoulders. Finally, the head and shoulders in the video are detected by the soft cascade classifier. The algorithm is programmed using C++ language, and selects traffic intersection videos for experiment. Experimental results demonstrate that above 90% detection rate is achieved, which meets the design requirements.