The Design Of License Plate Recognition System And Research Of Character Recognition Based On Improved SVM

With the development of national economy, the amount kinds of vehicles have increased rapidly. Although makes in deed our life convenient, it cause a lot of traffic surveillance problems. Intelligent traffic surveillance (ITS) has been the main measure and development direction of highway traffic and municipal traffic management today. In addition, car plate automatic recognition technique is the kernel of ITS. In this paper, we mainly discussed the design and implementation of vehicle license plate recognition system. According to integrative used image analysis and processing technology, artificial intelligence, pattern recognition technology, in summing up previous work and experience, the package and put forward a complete solution. Following work has been finisher in this dissertation:Researching the algorithm of license plate location. Against the high-resolution digital images in the background of complex have taken in the actual traffic junctions, in this paper, a wavelet-threshold denoising combining line-scanning methods of license plates for approximate location, and comprehensive utilization of license plate number of characteristics for the rapid positioning methods is given out. The locating methods make full use of the license plate of texture features, color characteristics and geometry characteristics, and the introduction of the concept of confidence to filter suspicious region. After the license early location, based on the Radon transform the license tilt angle detection, and precise locating license, we can fast and accurate locating to the location of license plates through three steps. The experimental results show high recognition rate and fast speed for high-resolution digital images in the background of complex, applicable to a variety of complex background of the actual license plate recognition system applications.Researching the segmentation algorithm of license plate. This paper proposes the use of decision of tree structure in the color space, to determine the type of license plate. Then according to the rules of character projection form space area vector, the characters in the license plate area are accurately split through the search with the rules of vector. This method can availably resolve the problem of how to segment the conglutination or incline characters correctly and the characters in the license plate area can be availably split.Researching the characters recognition algorithm of license plate. This paper proposes a half-integral projection from the license plate characters feature vector can effectively improve the character of the distinction between the characters feature vector and implementer the part of training and recognition of the plate license recognition system by the Support Vector Machines(SVM). By the means of SVM, better recognition rate could be acquired in the case that relevantly less training samples has been sampled. We research the SVM multi-class classification algorithm applied to the training and recognition of the plate license. The characters feature vectors are trained using a Sequential Minimal Optimization (SMO) and then character types are determined by the multi-class classification of one-against-one method voting.