A new method of vision-based seat belt detection. (2nd December 2019)

• Record Type: Journal Article
• Authors: Yang, Zhongming; Xiong, Hui; Cai, Zhaoquan; Peng, Yu
• Is Part Of: International journal of embedded systems. Volume 11:Number 6(2019)
• Page Start: 755
• Page End: 763
• Publication Date: 2019-12-02
• Subjects: seat belt detection -- connected components -- big data in traffic -- structured image data; Embedded computer systems -- Periodicals; 004.16
• Languages: English
• ISSNs: 1741-1068

Abstract

In the traffic management system, it can greatly improve the management efficiency if the algorithm that monitors automatically detects whether the driver fastens the seat belt; however, currently prevalent detecting methods cannot achieve satisfactory results in aspects of the detecting rate, the image quality requirement and the colour difference between seat belt and the surrounding environment. We propose a method of seat belt detection based on visual positioning. The algorithm locates the window according to the licence plate position and the contour statistics obtained from the gradient. The face detection is used to adjust and determine the seat belt detection area in the window. Finally, the method of seat belt detection based on the connected area is used to detect whether the seat belt is fastened. Experiments show that the successful rate of the proposed method is much higher than other existing methods, and satisfactory results are obtained.