Research And Implementation Of Gesture Detection Based On Video Image

Now the Video Supervisory System is widely used in people’s works and normal lives, with the development of the technology and the improvement of people’s living standards, the traditional Video Supervisory can hardly satisfy the people’s requirements. The spread of the Video Supervisory System led to amounts of surveillance video, traditional Video Supervisory which needs artificial processing cannot meet the requirements of processing a large number of surveillance video quickly. In order to processing amounts of surveillance video quickly, we must solve the problem that traditional Video Supervisory needs artificial processing.In this thesis, based on the fact that human motion normally occurs at the motion area of the image, first we obtain the foreground objects, and then we realize the gesture detection system based on the chamfer matching technique in the area of foreground objects, and then get the human behavior information, thus the intelligent surveillance is achieved. The main content of this thesis is as follows:1. Gaussian mixture model is firstly built, and then we extract the foreground from the image, the next step is to realize the gesture detection based on the chamfer matching technique in the area of foreground objects in this image. In this way we narrow down the detection area, and then the detection speed is greatly improved.2.The traditional chamfer matching algorithm has to process amounts of data and the processing speed is slow, in this thesis we propose a method based on RANSAC algorithm to streamline the edge feature, with this method we choose the line segments place of the edge points to decrease the storage space and the calculation.3. With the problem that the chamfer matching algorithm is easily influenced by the edge of the background, we proposed a method that the chamfer distance is augmented with an additional cost for direction mismatch, and then we generate the three dimensional integral distance transform, in this way we can improve the efficiency and accuracy of gesture detection.