A novel computer vision system for intelligent traffic management

摘要

A new moving vehicle detection algorithm is introduced in this paper. The algorithm is a modification to the traditional region-based detection method of background subtraction. The object of interest is separated from image by adaptive threshold segmentation. The idea of adaptive threshold image, which is a 2-D matrix of threshold values, distinguishes this algorithm from others. The values are not only dependent of time but also adaptive to the location of pixels. It generates better result than the traditional methods we have used before. In the rest of paper, an Intelligent traffic system (ITS) integrated this algorithm is also presented. The system consists of following processing modules: (ⅰ) stem Initialization (ⅱ) Automatic vehicle detection (ⅲ) Region Correction (ⅳ) Shadow Handling; (ⅴ ) Region identification. Each of them is explained in detail.