Research On Fatigue Driving Characteristics And Parameters Extraction

Fatigue driving detection system has a larger application value, and is closely related to people’s safety and has been a research focus, facial recognition fatigue method based on pattern recognition and image processing technology has important theoretical value. Research in this area has achieved some results, but there are a lot of unresolved issues. This paper focuses on the study of facial fatigue characteristics and fatigue parameters, proposes detection and recognition algorithm based on the fatigue characteristics of the three fatigue parameters, as well as the three fatigue parameter access methods, and does related experiments. This article research facial feature detection is aiming to obtain fatigue parameters, more targeted, and design the real-time acquisition of fatigue parameters. The final proposes a multi-fatigue parameters of fatigue driving system design method, the study can be summarized as the following:(1) Propose a eye detection method based on the Hough transformAfter detecting face region using face detection method based on the the Adaboost, and separating eye area, proposes an eye detection method based on the Hough transform. Do Hough transform with given radius and non-maxima suppression method to detect candidate eye round, join grayscale information to determine the human eye. Based on the binocular relations priori knowledge, designe a half the image detection method for detection eye algorithm, and add a second detection method for detecting process. Reduce the image to different resolution to do experiments using detection eye method in this paper, based on the experimental results analyse the influence in the calculation amount and detection results by the image resolution reducing. Select the appropriate resolution to reduce the amount of computation and keep better detection effect. Experiments show that the detection of the human eye algorithm for non-shelter, non-closed eyes has good detection results.(2)Training AAM human eye templateThis paper selects of training AAM eye template to precisely position eye and extract feature information. Firstly designed training methods:Training Single people, multi-eye status, multi-gesture human eye template and select the appropriate training samples. Choosing and designing eye feature points labeling method of the training sample, the selection of the feature points can provide a large number of eye even header information. Using AAM algorithm to the train sample and build the eye template. Then design the the eye information extraction method to select the pupil points around to characterize the eye opening distance. Finally design the eye tracking method. Using eye detection method based on Hough transform in this paper to early position the eyes, match the AAM human eye template, track eye and detect whether eye feature tracking lost. If lost using eye detection method in this paper to do eye re-positioning. The AAM human eye template method in the precise positioning of eye, the eye tracking, and real-time extraction information of the eye obtained good results.(3) Design eye fatigue parameters PERCLOS real-time acquisition methodBased on the human eye feature detection method, design eye fatigue parameters PERCLOS real-time access method. Combined AAM template for tracking and positioning the human eye, the chosen template has real-time access to the pupil four corners of point to characterize the eyes open distance. According to the definition of PERCLOS, this paper designes in a eye blink process recording the time of eyes open distance less than the entire pupil distance of20%and the time of less than80%, value of comparing the two is PERCLOS. Design achieving the real-time PERCLOS value output, and fully characterize the changes in the eye in real-time. Related experiment of the proposed method achieve better results.(4) Propose a mouth detection and identification method based on Gabor and LBPThe mouth fatigue parameters has been detection yawning frequency mainly to get the parameters. For the purpose of getting this parameters research method of detection and identification mouth feature state.Using the positional relationship of the eyes and mouth in the face area segmentation mouth area, extract the red information images, multi-scale and multi-direction Gabor filtering then improved LBP operator to process the image, and then using sub-block histogram extract feature, and design SVM mouth status classification based on judgment the non-fatigue and fatigue. Use of the results of the classification and identification, design real-time access to the mouth fatigue parameters, a period of time the mouth state continuous open record once yawn action.(5) Propose fatigue detection system method design based on this paperPropose A multi-parameter system of driver fatigue design method based on three fatigue parameters and their corresponding feature extraction methods. Based on links between parameters and their corresponding feature extraction algorithms, mainly research and design modules and process of system.