Research And Implementation Of Gait Recognition System Based On Video Stream

Due to the breakthrough of computer vision and artificial intelligence technology in the recent ten years, biometrics recognition attracted wide attention again. In the biometrics used in the identity recognition,Gait as a feature which fuse inborn and acquired character,not only has a high individual differences,but also satisfy a range of advantages such as non-intrusive?hard to disguise and low image resolution requirements.That makes gait recognition becomes hotspots in the biometric technology.Gait recognition can divided into three part—gait video preprocessing, gait feature selection and the recognition of the feature. In this paper, lots of experiments have been make.We perfectly illustrate the feasibility of gait analysis to identify the individual in theoretical and practical.Innovation of this paper is as follows:(1) In gait video preprocessing, we abandon frames subtraction, proposed pixel value cluster with no parameters in gray space combined with connected component labeling technology. Experiment shows that the extracted contour image is very clear which has an important significance in the following work.(2) In feature selection stage,aimed at the shortcomings in existing method of gait feature extraction,presented a gait descriptor based on body centroid and key points of outline.Effectiveness of the feature has been proved.(3) In feature recognition stage, improved automatic time warping algorithm in gait cycle and step pattern selection,make it more suitable for gait recognition, quantify the degree of different gait sequences.(4) Implement a gait recognition system based on the video stream.It use MFC + MySQL+ OpenCV as development environment.Take gait video as input, output the identity by using the algorithm mentioned above.Proved the feasibility and practicability of the proposed program,and laying the foundation for future applications in more complex situations.