Fast Car Logo Recognition Algorithm

With the increasing number of traffic accidents, monitoring becomesthe main technical point in the intelligent transportation system. The purposeis to get the license and the logo of cars, so as to investigate the aimed carrapid. There aremany monitoring technologies, such as the first generationanalog video surveillance system, the second generation of digital videomonitoring system which is based on PC+multimedia card, the thirdgeneration which totally based on IP. According to these monitoring systems,the key content is how to locate the logo quickly. Facing with the requirementof high speed, how to effectively position the logo is a significant task in thefield of pattern recognition research.The topic of the research is aimed at how to carry on logo’s positioningfrom a single image and make a match with original elements with theobtained various parameters. The researching content is divided into twopieces. A introduction is as follows:1. In view of the existing positioning method based on licenses, thepaper achieves the algorithm. Firstly, we need to preprocess the originalimage, in order to highlight the part of both the license and the logo that theycould be easily detected in the next step. Secondly, it is time to carry on thesegmentation. Based on the pixel of object is different from surrounding.Furthermore, the method of median and average filtering is also feasible.Thirdly, it is necessary to remove noises using the algorithm of morphology.Because there are noises during the first steps, therefore it is necessary tohave the step. Fourthly, in order to extract the license’s outline, we ought tocarry on the operator, such as Sobel and Canny. Finally, the logo could bedetected after using the relative position relation between the license and thelogo. 2. In the second part, we are going to do the research of the technologyof HOG and SVM. HOG is a feature of detecting target, and SVM is a kind ofclassifier. Referring to the algorithm of positioning of cars’ logo, it isnecessary to train the SVM classifier firstly. Then it is time to do matchingwith the library’s models with the value of HOG as the inputting, so that thelogo could be detected. The paper puts forward a new fast detecting algorithm,which is based on the current extracting model of HOG+SVM. Theinnovating point is that it could fast achieve the detection of logos.Considering the current demand of high speed, the algorithm is likely to beused in a wide range.The main content of the research is how to realize the logo’s positiondetection. The paper is involved in two plans. One is aimed at the maturetechnology of detecting cars’licenses. The other one proposes a new fast logopositioning detection algorithm on the basis of HOG+SVM.