Feature Detection And Analysis Based On Multi-view Of Facial Region During Futigue Drive

With the development of automobiles industry and road transportation, more and more people start to drive cars. As the increasing number of automobiles, traffic accidents happen more frequently, because of the fatigue driving of drivers. Now many researchers are studying how to detect the state of drivers on the load, it is an important method that use vision detection in this field. Although tactile detections on fatigue driving, such as electrocardiograph, electroencephalograph, involve the advantage of high detection rate, its complex procedure requiring much time and effort with different operations make it unreasonable for practical application. The detection with visual methods could give a more efficient extraction of eyes and mouths for the judgment of fatigue driving.A study is made on vision detections on fatigue driving, some results are got and the rotated angle are extracted from the detected area of face.At first, the knowledge about image pre-processing is introduced and then summary for the method in image processing. As the light imbalance in image acquisition, some light compensation for the target images are made.Then, the principle of Adaboost method is present. The author analyses its consistent and the procedure for face detection, especial for the selection of Haar characteristic and the computation of eigenvalue, and the classifiers. Then give some detection results of Adaboost.Once more, intrduce different color spaces, such as RGB, HSV and YCbCr. According to the clustering performance of face color in YCbCr and other color space, the area of face in the images can be extracted based on Cb and Cr quantities. And also the principle and extracted result of mouth and eyes are present.At last, Adaboost method and face segmentation based on skin color are used to detect the face in the picture. In order to get more accurate face area, date fusion theory are employed, which integrates Adaboost method and skin color segmentation. Furthermore, the rotated angles of face in images are extracted for the further use.