X. Zhang, H.L. Jiao, Y. Zhang, Z.K. Li, C.Y. Zhang

Pages: 157-168

Abstract
Accurate fatigue driving recognition is vital for road safety, preventing accidents caused by driver exhaustion, and protecting lives. To address the challenges of inadequate accuracy in driver feature extraction, limited precision in fatigue recognition, and prolonged processing time in conventional truck driver fatigue detection systems, the paper propose a novel fatigue driving recognition of truck drivers based on multi-feature fusion. Firstly, this scheme utilizes specialized acquisition equipment to accurately capture images of truck drivers, and corrects lighting inconsistencies through two-dimensional gamma functions; Then, the image is denoised using a dual neighborhood median filtering algorithm; Subsequently, the features extracted from the driver's eyes, mouth, and head were effectively integrated into an extreme learning machine framework to achieve accurate driver fatigue classification. Experimental results demonstrate the effectiveness of new approach, achieving a peak feature extraction accuracy of 98.74%, a maximum recognition accuracy of 98.56%, and a time efficiency ranging from 0 to 0.53 seconds.
Keywords: multi-feature fusion; truck drivers; fatigue driving recognition; two-dimensional gamma functions; dual neighborhood median filtering algorithm; extreme learning machine