Traffic measuring and authentication on digital video

Several schemes for traffic monitor on video are developed. The first is the dynamic gap technique by means of digital video processing instead of subsurface electromagnetic detectors. The dynamic background refreshing and threshold selection are obtained by statistical methods. The second is density and queue length measurement in a traffic detection region. The third is interpretation and identification of traffic activities by hidden Markov models and entropy minimization to discover the relationships between hidden causes and observed scenes. The learning system learns a concise model of scene behavior directly from optical flow. The fourth is to authenticate traffic images with robust digital watermarking technique and hiding data in image losslessly.