The Extraction And Effective Identification Of Human Motion Characteristics For Video Images

With the rapid development of information technology and popularization of the application, Using the computer vision technology to study image processing and pattern recognition, then to extract the human motion characteristics and to identify the person effectively for video images has become a hot problem. The identification of the body motion for video images with computer visual technology is analyzing and dealing with the video or sequence of images. It can detect the human motion target, extract the motion characteristics in order to understanding the movement and identifying the person. The human body movement analysis for video images has wide application prospects such as intelligent monitoring, intelligent interface and virtual reality, etc.The extraction and identification of human motion characteristics needs combined with biological recognition technology to identify and judge the behavior of human and identify a person’s identity. Biometrics recognition technology uses human biological characteristics to identify person’s identity. The biggest characteristic of that is the uniqueness and invariance property.Human motion characteristics include:the body swinging characteristics, gait characteristics, the human body outline projection characteristics and the human body symmetry characteristics, etc. Form the view of video surveillance, gait characteristics is the biometric with the most representative and typical in the case of long-distance. In recent years it has been paid more attention to, and a large number of meaningful gait recognition algorithm have appeared.Motion characteristics identification includes moving target detection, motion feature extraction, classification and identification. Based on the study of all kinds of human motion characteristics identification algorithm, the author has done the detailed study about feature extraction and classification. It has proposed a new improved algorithm in this paper. The main achievements of this thesis include:1. Extracted effective human motion characteristics.Because of many kinds of human motions and the impact of external factors, the recognition rate based on a single feature is very low. Using the simple human walking movement as the research object, it has proposed a simple effective algorithm based on fusing the static and dynamic features of the walking body to identify the identity in this context. The human body contour has uniqueness and is a static characteristic of the movement. Meanwhile people’s walking behavior includes hundreds of simultaneous movement of limbs and bone structures. The feature of limbs angles for motion information which is easy to fuse with other characteristics can describe movement characteristics. Based on the analysis above, a new recognition algorithm with the fusion of seven Hu moment invariants of the human body contour and the feature of limbs angles for motion information is presented in this paper to identify human identity together.2. Research about classification and recognition algorithm.This text aims to use the human body contour and the feature of limbs angles to identify people’s identity. However different feature have different data types and scales, a nearest neighbor fuzzy classifier is introduced. This kind of classifier is actually a simple fusion of multiple classifiers. Considering that different feature has different contribution on the result in the recognition process, we distribute different weights to different membership degree and use a kind of decision algorithm to get the recognition result. It can improve the recognition rate effectively.Experiment results show that the proposed algorithm has better recognition performance.