The Feature Extraction Of Vehicle Face Image

The difficulty of extracting features of the vehicle face is that the image presents a big difference in gray district due to the random reflect in the same physics area, consequently, the outline of vehicle face is so easily disturbed by the fake edge. To start research in analysis of the structure of vehicle face, this thesis proposes a feature extraction algorithm of vehicle face image.Firstly this thesis overviews the existing vehicle type identification technology in Intelligent Transportation and selects structure identification of saloon car face as the research content. Secondly we analyzed the general character and speciality of all sorts of car headlight and radiator gridiron in shape, dimension and structure. Lastly we put forward a description scheme and extracting frame of the car face features.This thesis studies the car face division methods based on the vehicle license plate localization. To begin with, we split the vehicle license plate from the car face image, then we calculate the vertical axle wire and the level cut-offline of the car face according to the size of the vehicle license plate.We have done a lot of works on outline detection method which is applied to the situation that the gray is non-homogeneous and random disturbance lineation region exists. The threshold segmentation and the edge detection method can not segment the headlight region accurately owing to the surface of the headlight is non-light tight and non-homogeneous reflection. Directed against this problem, this thesis proposes a headlight region detection method based on the Hough Transform and the Snake model. To start with, we detect the edge of the headlight image, next we deal with the edge by straight line detection, in addition we get the initial outline with a algorithm of 2-D convex hull based on included angle, finally we detect the headlight region accurately using the Snake method.We study the feature extraction algorithm based on the shape of the radiator gridiron, the structure of the gridiron, and proposes a algorithm of point-match radiator gridiron based on Hough Transform, a algorithm of radiator gridiron analysis based on horizontal projection and vertical projection. Besides this algorithm is used on many kinds of car's radiator gridiron . The test proves that this algorithm is effective.