The Research Of Embedded Driver Fatigue Detection Based On Image Information Fusion

As the increasing number of vehicles, it is particularly important to detect and avoid fatigue drive efficiently to reduce traffic accidents and casualties. Since the embedded system has features of easy-deployed, high reliability and low-consuming, the embedded drive fatigue detection system is becoming the main trend of future drive fatigue examination system. Therefore, in view of a conclusion and research on the drive fatigue detection technology in domestic and abroad, this paper presents a framework of embedded driving fatigue detection system which mainly focuses on eye detection and fuses head movement information. Moreover, this paper systematically studies the key technologies of multi-information fused face-detection, location, tracking and identifying of eyes, image information fusion fatigue state-identification and so on.The major work and creative results are as follows:(1) For skin color-based face detection method, this paper provides a skin color segmentation method based on YCbCr color space. It firstly makes illumination compensation on image, according to various light conditions. Secondly, it sets up the skin color model, and obtain skin likelihood image. By the application of self-adaptive method, it is feasible to separate the skin color area and determine the face region.(2) Based on the improvement of AdaBoost algorithm, this paper carries out experiments of detecting face by way of fusing skin color information, and then locating face precisely with geometrical condition as restrict. The experimental results show that the information fusion-based detection algorithm not only guarantees the detection speed, but also improves the accuracy, which satisfies the accurateness and real-time required by embedded fatigue examination system.(3) Regards to the extraction of fatigue features. This paper proposes the method of locating and tracking eye by combining projection and Kalman filter in the precisely located face area, and also proposes identification of the eyes state by using its corners feature. Secondly, according to acquirement of head fatigue information, this paper proposes a detection method of skin color proportion filter, to judge whether the fatigue information of doze and bow, based on which a standard to comprehensively judge the situation of driver's fatigue by integrating the eye fatigue and head movement information is proposed. The accurateness rate and reliability is verified .through experiments.(4) After analyzing the framework of the entire drive fatigue detection process based on Qt/Embedded development environment, the GUI (Graphical User Interface) and experimental results of drive fatigue detection system are given under the embedded development environment-built up by this paper.