Automatic road detection for highway surveillance using frequency-domain information

Video detection is one of the primary collection means of traffic states in Parallel traffic Management Systems (PtMS). In order to accurately and automatically obtain road areas in highway surveillance videos, this paper presents an automatic detection algorithm based on the frequency-domain information of video images. This algorithm uses the frequency-domain feature that is produced by the vehicles passing through road areas in videos, to realize automatic segmentation and recognition of the road areas. The experiment comparing with the traditional vehicle-tracking-based method, which uses the information in the time-space domain, illustrates the advantages of the proposed algorithm.