Research On Fatigue Driving Monitoring Technology

Driving fatigue caused by many factors, has become one of the most important causesof traffic accidents. It is of great significance to real-time monitoring and timely warningon the driver to reduce traffic accidents by using advanced technology. The driver fatiguemonitoring method based on computer vision, has been widely used in the fatigue earlywarning device, and become a hot research direction of fatigue detection algorithm by thecharacteristics of non intrusive and high reliability. Because of the strong correlationbetween eye behavior and driver fatigue, fatigue warning method based on eye behavior,has been occupying an important position in the field of computer vision method.Aiming at the three key problems of robustness, accuracy and real time in fatiguewarning method based on eye behavior and starting from face detection, eye locationfatigue decision mechanism, a new fatigue monitoring algorithm based on eye behavior isgiven. Its results are verified by experiments, the main work is as follows:1. Aiming at the problems of effects of background of face, head poses andillumination variation, a multi pose face detection algorithm based on the idea of tracking isproposed. This algorithm improves the correct rate and the speed of face detection. Anadaptive image segmentation method is proposed, which overcomes the effect of light onthe eye location.2. An algorithm of eye location based on the fusion curve and gray projection isproposed by improving the gray projection algorithm that be widely used in eye locationdomain, and lots of experiments is done in static face database to demonstrate the algorithmeffectiveness. The problem of gray projection and image segmentation in video sequence isanalyzed. On this basis, simple binary method is used for eye extraction, image momentdescription and context analysis is used for eye verification and optimization. At last, eyelocation and eye height extraction is realized. Eyes location and state analysis method inthis paper has a high accuracy at the same time a good real-time performance.3. Single fatigue recognition indexes have limitations in fatigue recognition, so weextend the sampling time to improve the accuracy of PERCLOS index method for fatiguerecognition, and time nictation index is fused with the PETCLOS index. A dual index threelevel fatigue recognition mechanism is design to improve the recognition accuracy offatigue. Dangerous action is considered, and through the face and eye localization success or not to determine whether there is danger action driver.In this paper, the whole fatigue detection system and the various modules is beentested, and the test data is analyzed to validate the fatigue monitoring system availability.From the testing results, the system can warn accurately and timely, has a good result. Thealgorithm use C++and OpenCv to program, that creates a good condition fortransplantation and system miniaturization.