Research And Software Design Of Driver Fatigue Detection

Recently, car is becoming more and more with fast urbanization process, the industry of logistic transportation is rapidly developing, and the sum of drivers is more than ever. At the same time, the road traffic accidents are also more and more. Studies have shown that fatigue is one of the important reasons for the increasingly serious traffic accidents. Therefore, further study of technology related to fatigue detection is very necessary, and has great social significance.Currently, fatigue driving detection based on image processing has such advantage as non-contact, high speed and strong anti-jamming capability, which draws the domestic scholar’s great interest, and it quickly became the hot research. As the constant progress of detection technology, people have higher demand for detecting accuracy.In this paper, some of the algorithms is studied in depth and make improvements,and we design the fatigue driving detection software based on the Android system. The concrete content is as follows:First, an eye contour fitting method based on adaptive particle swarm and least squares is proposed. In least square ellipse fitting, all the edge points are involved in the fitting process, and the existence of some larger deviation of edge points, as a result,the fitting accuracy is decreased. Aiming at this point, in this paper, the algebraic distance from Point to ellipse is selected as objective function, and we set up optimized ellipse fitting parameters model.Then we use least square to obtain the initial ellipse parameter values based on the optimized edge points.Finally we acquire the optimal approximate ellipse through loop iteration by optimized particle swarm model.The experimental result show that the improved algorithm has higher fitting accuracy.Second, we further study tracking model of the kalman filter and Mean Shift algorithm.To solve the problem that Mean Shift algorithm lost frames when the eye move fast, this paper proposes the method based on kalman combined with Mean Shift tracking algorithm and nonlinear kalman filter tracking model is set up. First of all, the center location information of the eye which changes over time as the observation value of the nonlinear kalman filter, nonlinear kalman filter is used to predict the possible positions of eyes in the current frame, and it is used as the original position of the Mean Shift, which reduces the search area of the target,and achieves satisfactory effect. Then, according to the gray level distribution characteristics of the eyes, use Mean Shift iteration algorithm to search template with eyes for the most similar goal in the estimate of the neighborhood. The simulation experiment shows that this method has a good tracking performance.We first verified the algorithms based on Visual Studio 2010 and Open CV. Then, regarding to the realistic need, we developed the fatigue driving detection software Based on Android, and test result show the software achieved the effective detection of fatigue driving. In addition, the subsequent function development of the driving detection system can also be based on this software.