Analysis Of Human Motion In Video Image Sequences

Analysis of human motion is one of the most important research topics in the domain of computer vision. It is also an active field that has interested many researchers in recent years. It has vital applications in physical exercise, rehabilitation process, virtual reality, robot, security surveillance, etc. Human motion analysis aims at attempting to detect, identify and track human bodies from a set of image sequences including human body, and more generally, to understand and describe their behavior.The object researched in this thesis is video-based human motion analysis in terms of detection, tracking and analysis of human motion in video. In order to detect and track human motion, this paper follows two steps:how to detect human body joints and get their position information in the image sequences; and how to track the human body joints in the next frame after detection. With digital video Processing and digital image Processing, this Paper mainly Proposes the solution ideas for human motion video:1. With studying the background initialization of background subtraction, this paper raised a method of background subtraction, which based on Change Detection Mask to structure background model. It proceeds to segment the moving objects by calculating the difference between the current and background. Obtain the motion target and update the background model in real-time.2. It achieved marking feature points of body joint automatically form a binary motion image. First, after extracting motion human's edge by Canny's detector arithmetic, we can get a appropriate shape of human body contour through Douglas-Peucker vector compression algorithm. And then found out the position and marked the feature points of every body joint according to the proportion of human body joint.3. Researching on the technique of 2D human motion tracking, this paper implements optical flow algorithm to tracking the marked feature points. Any wrong tracking result will be corrected by Kalman filter. The method of automatic joints marking of human body model can reduce the manual work in the system.