The Algorithm Of Vehicle-Logo Self-Recognition System

With the increasing of urban traffic pressure, the importance of intelligenttransportation system has been increasingly revealed. Since the currentvehicle identification system has been widely used in many fields, in thegarage management, road charges, all captured and accident treatment play animportant role. Vehicles since the realization of the recognition system hasgreat economic value and practical significance.Logo position and vehicle-logo recognition are two key technologies inthe vehicles since the recognition system. In this paper, we study a kind ofalgorithm implementation since traffic sign recognition system. The firstchapter, elaborated the research background of the subject, the developmentof intelligent transportation systems at home and abroad was introduced, andthe model of vehicle identification technology included in the recognition,license plate recognition and made a brief introduction of the vehicle-logorecognition technology, and analyzes the research of this topic logorecognition system development present situation and technical difficulties.The second chapter, we studied the basic theory of vehicle-logo location andrecognition. In terms of logo pretreatment was studied using weightedaverage method of digital image gray processing; Through histogramequalization to enhance contrast of the method; Use of neighborhood averagemethod for smoothing denoising method. This topic using edge detectionmethod for logo detection and localization, commonly used in classicaloperator method are introduced first, Roberts, Sobel edge detection operatoredge detection operator, Laplacian edge detection operator and the theory andapplication of Prewitt edge detection operator, then the optimal operator is introduced in this paper the Canny operator, LoG operator in the theory andapplication of two operators. Vehicle-logo recognition part is the first tointroduce the badge feature selection principle and four commoncharacteristics: the texture, color, space relation and shape, and thenintroduces the principal component transform and Hu two invariant momentfeatures transform method. The next decision describes the minimum distanceclassifier for classification, bayesian classifier and support vector machineclassifier and neural network classifier four classifiers. The third chapter, thesystem are discussed in detail logo positioning algorithm based onbackground suppression and morphology filter implementation. Using themature first license plate location algorithm and the prior knowledge, torealize the coarse position of logo; Then through weighted average gray levelmethod and the neighborhood averaging method to realize the logo image andvalue smoothing denoising; Using Sobel edge detection operator to logofurther testing; Then through background after denoising, the use ofrectangular structure element can realize accurate positioning logo. The fourthchapter, the logo first moment feature is extracted, and then according to thetorque characteristics of logo, select7Hu moment invariants, as recognitionfeature invariants are classified, then through established good standardcompare the feature library, using Euclidean distance to do logo similaritymetric finally realizes the vehicle-logo recognition. Through a large numberof experiments, proved that the identification algorithm with high precision,fast processing speed. The fifth chapter, the study of this article made asummary, for the current list the advantage and disadvantages of the system,and the improvement of the system were discussed.