A Human Detector Based on Flexible Pattern Matching of Silhouette Projection

In this paper. we propose a method for detecting a human being using silhouette projection pattern matching, and present experimental results of its application in video surveillance systems. There are some studies on detecting intruders in video images by exuacting changing regions which follow moving targets. However, detection e m caused by small animals, changes in brightness due to sunshine, or the swaying of bws in the wind cannot be avoided. Several methods besed on additional constraints such as the area or the rectangle surrounding the changing region are proposed, but they are not sophisticated enough to overcome the above problems. Our method is bawd on silhouette projection pauern matching of the changing regions to reduce erroneous detection. Although the shape of human silhouettes change as the person moves, the widths of certain parts of the body, which are elements of the projection, are stable. Therefore, we propose a flexible pattern matching algorithm where different allowance levels are set to different parts of the body. This method is simple and takes short calculating time. The algorithm was implemented in specially designed hardware which was able to extract and evaluate the changing regions in real-time. The experimental resultq prove the effectiveness of this method.