Image-based seat belt detection

This paper presents a novel method for the detection of seat belt in a monitoring image which contains the full scene information of the moving car. First, the driver area is located based on the vehicle outline. Then the potential seat belt edges are detected by an effective algorithm based on the direction information measure in the HSV color space, and the result is finally obtained by further verification of the edges. Experiments demonstrate the method makes a good performance even with noisy images.