基于视觉静态交警手势识别算法的研究-控制科学与工程专业论文.docxVIP

基于视觉静态交警手势识别算法的研究 基于视觉静态交警手势识别算法的研究 VISION BASED STATIC TRAFFIC POLICE GESTURE RECOGNITION ALGORITHM RESEARCH ABSTRACT With the development of computer vision and image process technology, human motion recognition because of its wide range of applications, now become a major concern of the research in the field of computer vision. Vision-based recognition of peoples motion not only includes the knowledge of image process and computer vision, also involves the theory of pattern recognition and artificial intelligence. This paper studies the traffic police gestures recognition and hope that it will be used in driver assistance systems. Traffic police gesture recognition algorithms are compared and researched by a large number of experiments. The previous histogram-based image segmentation is often concentrated on the grayscale image. In this paper, color image are segmented by the color histogram based on the RGB color space, so it overcomes the shortcomings of using histogram segmentation. In order to effectively suppress the interference of external factors such as peoples clothing and overcome the current shortage of the whole model identification algorithm, we present a feature extraction algorithm by combining characteristics of human body shape and human skeleton and use Euclidean distance-based template matching method to identify. In the recognition experiment a template library with 8 kinds of traffic police is set up. Then research on recognition algorithm with a traffic police gesture set of 771 samples created by 6 people. The average recognition rate reaches 84.95%. The method is not sensitive to interference of perspective and clothing. It also has a good robustness. Experimental results show that body shape can better reflect the characteristics. The method has a higher recognition rate that combining characteristics of human body shape and human skeleton. KEY WORDS: computer vision, motion recognition, feature extraction, traffic police gesture ii 原创性声