VEHICLE DETECTION FOR TRAFFIC MEASUREMENT USING EDGE AND BACKGROUND IMAGE

Automatic traffic measurement systems, such as magnetic loops and ultrasonic sensors, are inflexible in terms of modifications and additions. A new algorithm for detecting vehicles from traffic images obtained from mounted cameras is presented. Since background images (BGIs) are advantageous in distinguishing vehicle from background, and edge detection is stable under fluctuation of the illumination condition, the algorithm creates a BGI and a background edge mask (BEM) adaptively and then extracts vehicle edges using BEM. Extracted edges are projected horizontally for every lane, the projections are adaptively binarized reducing noises and filling holes, and the vertical location of vehicles is detected from the projections. Moreover, the algorithm is tested by computing traffic volume of an actual traffic scene using a method that tracks vehicles and recognizes their passage.