Some Applications On License Plate Recognition Of Digital Image Processing

This paper applies several mathematical methods to focus on studying several image processing technologies, such as image denoising, image super-resolution, moving object detection and character recognition, etc., and carries out the research and development of License Plate Recognition.Chapter 1 mainly introduces the origin, research significance and related review of our paper, and briefly describes our study.Chapter 2 proposes a difference scheme based on nonlinear diffusion Perona-Malik model for numerical calculation in image restoration. Our scheme can adapt to determine the tangent directions to the isophote lines based on two mutually orthogonal directional derivatives, which results that diffusion is along the edges as much as possible. One of typical edge stopping functions for Perona-Malik model is modified in order to improve robust calculation and satisfy the compatibility, stability and convergence for our numerical scheme. Computer experimental results indicate that the algorithm corresponding to our numerical scheme is very efficient for noise removal in regardless whether the noise is serious or not.In Chapter 3, a new kind of image super-resolution method is investigated. As far as we know, it has attracted widespread attention that how to get a high-quality high-resolution image as fast as possible.Based on properties of Legendre polynomials, we first establish piecewise bilinear polynomial wavelets system used to approximate an image function which is related to gray-level values. These wavelets are handled by equidistant sampling and Gram-Schmidt orthogonalization, so that they can precisely express the digital image by adopting linear approximation of discrete domain. And then, according to linear combination, a super-resolution image is generated from their more dense resampling. We further obtain higher-quality super-resolution image by multidirectional staggered constructions. Finally, the induction of matrix operation makes that our desired image is practically produced in real time. Compared with some excellent techniques, numerical and visual results of computer experiments indicate that our method reproduces more faithfully the low-resolution image in high-resolution form.Combined multiscale partitions of image space and 3-frame difference method, Chapter 4 proposes a moving object detection model for the high-definition digital video in a complex traffic scene. Firstly,key frames of being used to detect are extracted at regular frequency from the traffic video. Secondly, the gray images of these key frames are segmented by multiscale partitions. Based on the average gray value of every segmented window, we adopt 3-frame difference method to detect the window areas of moving objects for key frame images under the same scale partition, and further to integrate detection results at different scale partitions. Next, the method is implemented by matrix operation on the machine in order to complete real-time detection of moving object for the traffic video. Finally, we utilize some traffic videos which were captured in different environments and scenes to test our proposed model by programming on Matlab. The results indicate that it can real-time and excellently complete detection tasks in a complex traffic scene.In order to utilize the template matching method more efficiently to implement the recognition of license plate characters, we propose a fuzzy template matching method based on mathematical morphology which combines mathematical morphology and fuzzy set theory in Chapter 5. Firstly, for each binary image pixel point and its 8-neighborhood, the degree that the center pixel point belongs to the current character can be described by weighting. Secondly, we use a 4 × 4 window to select a representative point and traverse it through the entire character with partially overlapping, and the fuzzy membership matrix of each character image is constructed; Then Hamming approach degree calculation is utilized to classify the characters and the recognition is achieved; Finally, we program the fuzzy template matching method on the Matlab and test the recognition results of actual license plate characters. Compared with traditional template matching methods, the test results show that our proposal can improve the recognition accuracy of license plate characters significantly.In the research foundation of previous chapters, Chapter 6 integrates related technologies of image processing and develops graphical user interface of License Plate Recognition on Matlab. The interface focuses on design of the interactions between user and computer. It mainly implements several functions,such as reading video files, setting parameters of detection, detecting moving objects and recognizing license plates, and saving results, etc.