Research Of Driver Fatigue Detection Based On Information Fusion

Fatigue driving is one of the major causes of traffic accident,and driver fatigue detection has become hot spots of intelligent traffic regions.This thesis first summarizes the current research of driver fatigue detection and explores the method of fatigue detection based on information fusion:firstly get the facial information using a camera,then obtain eyes,mouth and fatigue-related feature information through the corresponding feature extraction algorithm,finally integrate multi-source information by the constructed D-S evidence theory and fuzzy neural network model to complete the judgment of driver's fatigue.The core study of this thesis include:research and realization of human face location algorithm.eyes and mouth location, characteristic index extraction of fatigue, construction of Information fusion model and realization of fatigue detection system.About the construction of model,this thesis firstly elaborates the ideas of information fusion in the field of fatigue detection,respectively makes the appropriate analysis to feature level and decision level,including the standardization of characteristics, the determination of characteristics index weights,the construction of decision-making model,the selection of appropriate algorithms and other issues.Then construct the fusion model of feature level combined with weighted average method and the fusion model of decision level combined with D-S evidence theory and fuzzy neural network.Experimental results show that the fusion model is well established to achieve the desired results, which effectively made up for deficiencies brought by using only a single source of information.About fatigue information extraction, this thesis firstly locates the face combined with cascade classifier based on haar features.Secondly, track and locate the eyes with average of synthesis exact filters algorithm, and on this basis,obtain the approximate location of the mouth combined with "three stops five eyes" algorithm.Then consider the percentage of black pixels to all the pixels in the framed area of eye as the measure of the degree of eye closure,and the difference of the degree of eye closure between two adjacent frames as the closed speed of eyelid in the single time.Finally, in aspects of system implementation, obtain a better result through the algorithm simulation under laboratory conditions. In addition, this paper adds the storage and comparison function of historical data, which effectivly avoids the individual differences among different drivers,greatly improves the accuracy of fatigue detection.