Real-Time Counting Method for a Crowd of Moving People

Currently, the video processing is broadly adopted to deal with the moving crowd counting issue in many literatures, and an acceptable accuracy was obtained under the situation that the crowd is loose, on the contrary, the accuracy will decrease. The method with higher accuracy, even under crowded situation, was proposed. We film the crowd people while them entering or leaving, a camera was mounted vertically above them. Then, the proposed algorithm of motion object detection and segmentation is applied to the film. It applied the frame difference algorithm based on the morphological processing to solve the light changing problem. According to the rate of height and width, we can solve the crowd overlap problem, and the accuracy of our algorithm increases.