Learning-based seat belt detection in image using salient gradient

Seat belt detection is a critical challenge in traffic video surveillance. To tackle this challenge, an automatic seat belt detection approach is proposed in this paper. The proposed approach yields three significant steps. First, an edge detection method is used to extract edge features of the input image. Second, a salient gradient map is constructed by considering salient gradient features in the image. Finally, the obtained salient gradient map is incorporated into a machine learning approach to perform binary decision on whether the input image yields seat belt or not. Experimental results are provided to demonstrate the superior performance of the proposed approach.