Research On Human Motion Recognition Based On Video Sequences

Vision monitor on motive object is a burgeoning subject in the field of Computer Vision. It distinguishes to the traditional monitor and control systems in its intelligence. Computer in the system not only like people's eyes, but also substitute people and help them to complete the task in short. So it could abate the burden for human beings, save the money, manpower and material resources, and the more important is it could discovery the dangerous and avoids crime in time. For its wide application foreground and great potential economic value, it is attracting more and more interest of famous research institutions and researchers all over the world.There are three important approaches in human motion recognition, include checking, tracing, and recognition moving targets'behaviors. Checking and tracing the targets are foundation of behaviors recognition, so in this paper introduce the two steps briefly and emphasize research the behaviors recognition. In the beginning, the base image operation library and the methods are introduced, and the popular methods in checking and tracing moving targets are enumerated. In the end, expound the method in this system after comparison.In this paper, investigate and finally find out the method aiming at the human motion recognition. Hu square and Hidden Markov Models are considered, the experiment result shows that the Hidden Markov Models have a good performance but also have some disadvantages. Finally, an improved method which combines the Hidden Markov Models and the Convex Hull Algorithm is put forward. The experiment data show that the improved algorithm has a good capability.