Research On Key Technology About Driver Fatigue Detection Based On Fusion Facial Features

With the development of the automobile industry and transportation, traffic safety has become an unavoidable issue. As a serious factor which threats traffic safety, driver fatigue caused huge casualties and property losses to the country and society. Therefore, it is important to design a real-time, accuracy, and robustness driver fatigue detection method for protecting drivers and passengers.Based on the principle of driver fatigue detection methods proposed by predecessors’ studies, this paper carried out some research work aiming at the problem of how to improve the accuracy and reduce the time for fatigue judgment, and proposed a driver fatigue fusion detection algorithm based on multiple features. The method firstly used a camera to capture the facial image of driver in real-time, using active shape model to locate facial features, and then extracted the eyes and mouth features that can directly reflect fatigue status. Finally, we put forward a fatigue detection algorithm based on fuzzy inference system. The algorithm can estimate the level of fatigue smartly according to the experience of human, thus the fuzzy concept of fatigue can be quantified accurately. The main contents of this paper are as follows:(1) In order to improve the accuracy and robustness of driver fatigue detection algorithm which is based on a single feature, this paper proposes a multiple feature based driver fatigue fusion detection algorithm. Two facial features(eyes and mouth) that could directly reflect the fatigue are chosen to estimate the state of driver synthetically.(2) This paper compared and analyzed the methods that currently used in face location. Taking executive speed of the algorithm into consideration, we used a fast face detection method based on simple Haar-like feature cascading Adaboost to locate face range. Thus we can constraint the initial research shape and improve the convergence speed of ASM.(3) Studied a feature location and extraction method. Firstly, we used ASM to locate eyes and mouth features in the facial range. And then we extracted the parameters of height and width for eyes and mouth through calculating the distance between the eyes and mouth feature points.(4) Focusing on the problem that it is difficult to determine the three-level classification(awake, fatigue, severe fatigue) of driver’s fatigue level, we put forward a fatigue detection algorithm based on fuzzy inference system. The algorithm integrated eyes and mouth feature parameters can estimate the level of fatigue smartly according to the experience of human.