Research On Detection Method Of Dangerous Driving Behavior Based On Apparent Characteristics

With the development of social economy,road traffic safety has become one of the major safety problems,and the dangerous driving behavior of drivers is an important cause of safety accidents,threatening the safety of people’s lives and property.In order to solve these problems,image processing technology has become the most potential technology to identify the dangerous driving behaviors of drivers.In order to study the driver’s dangerous driving behavior,this paper starts from improving the accuracy and real-time of the detection algorithm,focuses on the two aspects of face detection and facial feature point location,improves the algorithm combined with the characteristics of the dangerous driving behavior detection task,and proposes a multi feature dangerous driving behavior detection method.The main research contents are as follows:(1)In view of the low real-time performance of the traditional face detection algorithm based on skin color and Ada Boost,a cache mechanism is added to store the size and location of the face,which reduces the detection range and improves the detection speed.A face location algorithm combined with particle filter is proposed.When the face detection algorithm fails to detect the face,the face location is carried out,which effectively solves the problem of face detection algorithm failure in the case of occlusion or deflection,and improves the accuracy of face location.(2)The algorithm of Active Appearance Model(AAM)is optimized.The Histogram of Oriented Gradient(HOG)combined with Support Vector Machine(SVM)makes initial classification of the driver’s head pose,which makes different face pose match different face initialization models,and reduces the fitting times of feature point extraction process.Based on the extracted facial feature points,the head posture is solved,and the driver’s distracted driving behavior recognition strategy is proposed.(3)A hand location method based on face mask is proposed to locate the driver’s hand,and the hand position is regarded as a factor to judge the calling behavior.The data set of driver’s driving posture is collected.Aiming at the defect of single hog feature in describing the target,the reduced dimension hog feature and Local Binary Pattern(LBP)feature are fused to classify the driver’s driving posture.Aiming at the problem of poor robustness of driver’s telephone behavior detection method based on visual features,a multi feature based driver’s telephone behavior detection method is proposed.The telephone behavior is determined by driving posture,mouth action and hand position with different weights.Finally,the existenceof telephone behavior is judged.Finally,based on MFC and Open CV open source library,a video oriented driver dangerous driving behavior detection system is designed and built.The experimental results show that the system not only has good real-time performance,but also can effectively detect the dangerous driving behavior of drivers.