Incident detection in pedestrian traffic using image processing
Junyu Yin · 1996
A real time implementation of some image processing techniques for estimating crowd properties has been developed. Crowd density can be estimated by using background removal or edge detection. An algorithm for moving edge detection has been developed to estimate the number of moving people which is simple enough for real-time use. The feasibility of incident detection in crowds has been demonstrated using these techniques with commercially available image processing products. The capability of the system has been successfully demonstrated in real-time using video recordings taken in a busy railway station concourse in London.