Research On Some Key Technologies In Fatigue Driving Detection System

Traffic safety is an eternal theme in transportation industry. In recent years, with the rapidgrowth of the vehicle, the incidence of traffic accidents is more and more high, and it bringsabout the increasing serious harm to human beings. In the severe environment, traffic safetyassistive technology has been got the widespread attention. As an effective method to avoidtraffic accidents and reduce the loss of accident, it becomes a research hotspot in the field oftransportation engineering, and represents the development tendency of future vehicles.The thesis carries on researching some key technologies of fatigue detection system.Specific research works are as follows:1. Face detection and tracking. Using AdaBoost face detection algorithm for face detectionand analysing Camshift algorithms purpose to achieve the process of tracking face.In addition,improving the shortcomings. It achieves the face tracking window automatic initialization byusing AdaBoost face tracking algorithm and proposes a Camshift face tracking algorithms whichis based on eyes template matching. The algorithms solved the problem of tracking errors in thescene which exist in large skin colors.2. Firstly, importing a new method of mouth feature extraction that is based on LBP-TOP(Local Binary Patterns From Three Orthogonal Panels). LBP is not only a kind of extraordinaryeffective operator of describing texture, which can measure and extract the information of thelocal field texture in gray image and capture little detail characteristics of the image, but alsobased on LBP and combines the temporal angle, which extracts the texture feature of the mouthimage through three orthogonal planes. The purpose is to express the substantial information ofmouth movements better.3. Bring forward a method of mouth feature extraction based on CBP-TOP, which improvesthe operator of LBP feature extraction, and introducing CBP(Centralized Binary Patterns)operator in order to further improve the feature extraction of the LBP operator which exist some flaws, and divided the three panels into two classes, which use LBP operator and CBP operator,and extracted the mouth feature through synthesizing the three features as SLBP-TOP(Synthesized Binary Patterns From Three Orthogonal Panels).4. Recognizing eyes and mouth fatigue. Classifying the mouth feature which was extractedby SVM classifier and making the judge rules of mouth fatigue through a lot of experiments, andcompleting mouth fatigue recognition. Recognizing the driver eyes fatigue based on PERCLOS,through the eyes location and eyes state detection.