Font Size: a A A

Study On Driver's Face Detection And Location

Posted on:2006-11-22Degree:MasterType:Thesis
Country:ChinaCandidate:H Y ChenFull Text:PDF
GTID:2168360155972390Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Investigation indicates that traffic accident in highway is rising because of fatigue. Then. how to detect and prevent driver's fatigue has practical significance in reducing the traffic accident and death rate. Face detection and location is the key issue in the driver behavior surveillance. In this paper. a detection and location method is proposed. which can solve face detection under rotation in image plane and random pose. based on color information. active contour and facial geometric triangle relationship. A coarse-to-fine location algorithm is introduced to locate driver's face. The coarse location algorithm is to detect face contour from the collection images. and the fine one is to get the exact location by Geometrical relationship of the key facial feature. Four major issues are studied: Firstly. an algorithm of skin color detection is studied that can be used under variant light conditions after a thorough analysis of light effect on skin color detection. Then. based on skin color pixel analysis static. an optimize method is introduced to optimize face candidate region by the ratio of height to width and proportion of skin -color. Secondly. in respect that the detection result using Snake algorithm is affected by the origin position of snake model. an improved Snake algorithm that combines region and grads information is proposed. The results of the computer simulation show the advantage of the proposed algorithm. Thirdly. the ellipse rigidity parameter and an ellipse detection method are introduced based on geometry shape after the research of common facial feature location algorithm and the character of elliptic black hole after skin segmentation. Finally. after introducing the simple verification based on face geometry shape. size and the ratio of skin color pixel in candidate region. a new verification based on facial triangle geometrical relationship is brought forward. In this paper. VC++ 6.0 is used to develop the general control software. realize various functions of the algorithm. and perform monitoring experiments. The satisfactory experiment results prove that the algorithm is valid.
Keywords/Search Tags:Face detection, Color segmentation, Snake algorithm, Ellipse detection
PDF Full Text Request
Related items