Vehicle Logo Location And Recognition Based On Image

Video analysis technology plays an important role in the field of intelligent transportation. As a part of the vehicle information, Logo has a broad application prospect in dealing with deck, fake brand and other illegal acts. Aimed at the low rate of logo positioning and the high rate of wrong recognition, this paper focuses on the logo positioning and recognition algorithm in the natural environment.License Plate Location Algorithm for a single feature information exists many problems, such as imprecise positioning and license plate error localization. This paper proposed a license plate location technology based on the combination of the license plate region texture jump and color space information. Firstly, the license plate candidate region was confirmed by using image preprocessing technology. Secondly, the license plate texture information could be used to determine the up and down, left and right sides of license plate borders and remove the false license plate area. Finally, the HSV color space information could identify license plate color and remove the false license plate area. Experimental results show that the algorithm improves the precise localization rate of the license plate and reduce the error rate of the license plate positioning.According to the vehicle logo location.Firstly, vehicle logo candidate region was determined based on the fixed spatial relationship between license plates and Logo. Then, on the basis of the correlation of the horizontal and vertical edges, inhibition operation on the edge background of the logo area was executed. By using area filling technology and connected components labeling technology, it could realize accurate positioning of logo. Experimental results show that the correlation detection method effectively inhibited the effects of heat sink and improved target localization rate of vehicle-logo.Recognition method of vehicle logo exists higher misdetection rate and low rate of error identification. This paper proposed a vehicle logo recognition method, which is the combination of HOG and ASIFT feature. Firstly, it extracted HOG feature by transformed logo samples and trained multi-classifier model which is based on probability classification by using SVM. Secondly, it extracted SIFT feature of each vehicle-logo and established vehicle-logo template library based on template feature fusion. Finally, vehicle logo recognition was completed by SVM multi-classifier and ASIFT match technology. Experimental results show that the size of training samples, number of samples, templates and other factors lead to low rate of average recognition. The implemented method has higher recognition rate than the single recognition method. Meanwhile, this method can effectively decrease the rate of error identification and reduce misdetection rate.