Monitoring And Analysis, Based On The Asymmetric Placement Binocular Camera Driver Attitude

In this article, our research content is making analyses and judgments to driver's head, shoulder, elbow and hand position and area through using two asymmetrical laying aside cameras ,which can predict the potential dangerous situation and prevents its occurrence. In the driving process, depth information of driver's posture doesn't change much, whereas its side information contains fich, thus in this article we have used two asymmetrical laying aside cameras instead of the traditional binocular vision, that is to say we lay straight ahead and side ones respectively in order to fully and comprehensively gaining the driving posture information.The content studied in this thesis can be mainly divided into four parts: segmentation, detecting, tracking and analyzing .In the first part, we first establish background model by taking pictures of background image. Then the foreground image with movement can be got via using background deduction method. In the second part, we first establish the skin-color model by means of statistical results of face samples. Then both anthroponomy and skin-color model are need to localize face and hands. As to shoulder and elbow localization .it's always a difficult problem. According to the restraint rules defined by ourselves and human body structure knowledge shoulders can be detect. We make use of projection and edge detection to detect elbows. In the third part, the mutual correlation criterion of template match method has been applied to face, shoulder, elbow and hand tracking. In the last. part, analysis to judging whether the driver is in unsafe posture has carried on in the light of the definition to illegal posture rule.This article has finally designed a test system for postures analyses. The system is simple and easy to operate; we don't need any contact involvements, only two ordinary cameras, a computer with common configuration is necessary .The experiment proved that this system can distinguish the unsafe driving condition and has good space and time performance.