Study On The Algorithm Of Driver Fatigue Expression Recognition Based On Image

According to an authoritative survey, in the frequent major traffic accident, the vehicle driver fatigue driving is usually the culprit of the accident, so real-time detection and identification of driver fatigue state, has an extremely important significance in controlling traffic accidents caused by driver fatigue.The first and foremost task of driver fatigue expression recognition is to detect the face area of the driver. Face detection is a key foundation work in the automated identifition system, and the detection accuracy directly impact on the subsequent recongnition results.After the face region is detected, analyzing the difference between the state of the driver’s normal state and fatigue, we find that the most significant difference is the degree of driver’s eyes closed. Then we extract the characteristics of the state of the driver’s eyes, whether as identifying the driver fatigue driving important basis.The main research contents of this paper are as follows:1. This article briefly describes the concept of fatigue detection, application and domestic and foreign research status, and then detection the driver’s face based on the improved AdaBoost algorithm, what’s more, it is a real-time detection of the area of the driver’s face.2. After finding the driver’s face region, we use the ASM model to locate driver’s eye, and then process the eye portion image by adaptive binarization, geting the binarized image of the driver’s eye.3. After geting the extracted binarized eyes, we can find the position of the eye through the ASM model, scan vertically the binarized image over the position of the eyes image, and this scanned gray values (binarized) is a feature vector, and then we scan vertically the binarized image over the positon that has a given step size (in pixels) in the horizontal direction with the position to obtain a two feature vector, and the last do exclusive-OR operation between the three feature vectors, then the result is the final feature. This feature don’t only describe the driver’s driving state characteristics, but also improve the robustness of the feature vector.4. We use the driving status characteristics extracted from the samples to train train the neural network, and then get the classifier to identify driver fatigue state.