| Vehicle recognition is a hot issue, with the increase of global vehicle, also brought more and more problems. Intelligent transportation become the field that each country wants to development, and the vehicle identification is one of the important applications.First the paper tells us some foundation knowledge about related technology. Image enhancement and recovery are the two aspects of image pretreatment, and their purpose are to improve the quality of the images. The features of image include color feature, texture feature and shape feature. Then the paper introduces the feature matching and distance sorting.The three methods to recognition vehicle in picture is the focus of this article. First taking Hog features get vehicle area.Than introduced two methods based on space distance of picture matching algorithm. Global feature matching is the first methods, and taking the brightness channel of the color models of RGB, HSV and YCbCr color models to convolution with schmid and gabor filter, and get the whole feature value set of vehicle image, and then used Euclidean distance to calculate the distance between the image and the image, and find out the minimum value. The second method is local characteristics identification that is cut of the image into many small pieces, the method mainly based on combine Lab color model with SIFT features, and get the new features of local space dColorSIFT, then use the Euclidean distance method to calculate the distance between the characteristics of each region,and get the minimum distance from the neighborhood,than add all the minimums,and get the minimum of the picture last. The third method is based on the Ranking SVM for identification the different vehicle images, mainly using the local characteristic values dColorSIFT, and rearrangement the images from the most relevant to uncorrelated according to Ranking SVM, and find the most relevant image. The experimental results show the effectiveness of the algorithm. |
PDF Full Text Request