Fatigue Detection Based on Regional Local Binary Patterns Histogram and Support Vector Machine

Abstract

Driver fatigue detection is a challenging problem in intelligent transportation system. The detection accuracy suffered by different illumination. This paper proposed a fatigue detection method based on regional local binary patterns histogram (RLBPH) and support vector machine (SVM). Firstly, we division the face image into blocks, then using local binary patterns(LBP) operator to present each block and calculating the LBP histogram(LBPH) of each block, then combine them into a RLBPH to present the face image. We used the fatigue face sample RLBPH feature and normal face sample RLBPH feature to train the SVM to get its model and the parameters. We input the RLBPH feature of the testing sample to the trained models, thus can classify the RLBPH feature of the testing sample. In our experiments we observe that RLBPH features perform stably and robustly on different illumination, and yield promising performance in low-resolution images captured from web cam.