Researches on Driver Fatigue Detection System,which aims to ensure the safety of operations and to reducetraffic accidents caused by artificial factors, has been themajor research subject in transportation safety. There is anenormous advantage in the method obtaining the driver'simage by camera, We propose efficient tracking anddetecting algorithm and with an appearance model based onhaar-like features, finding out the accuracy and robustnessof tracking of eyes movements and the conflict between realtimetracing and accuracy of fatigue detection algorithmssystems. First, PERCLOS algorithm is adopted to analyzeand determine whether a person is fatigue. Second,AdaBoost algorithm is applied to fast detect and thealgorithm is implemented in FPGA. Third, We propose acompressed sample tracking algorithm, which compresssamples of image using the sparse measurement matrix andtrain the classification online. The algorithms runs in realtimeand is implemented based on ARM add FPGAplatform. Experimental results show that the algorithm hashigh recognition accuracy and robust performance underreal train driving environment, in the case of nonlineartracking of the human eye, illumination change, multi-scalevariations, the driver head movement and pose variation.
展开▼
