Research On Video-based License Plate Recognition System

License Plate Recognition System (LPR) is a kind of system which make use of computer, image, video capture device and a number of ancillary hardware to locate and identify the vehicle license plates which is on the stationary or moving vehicle. LPR is widely used for research on non-site enforcement and monitoring such as automatic toll collection, traffic enforcement, security control, vehicle emissions. LPR is a core component of intelligent transportation systems.After vehicles are detected by testing equipment, through obtaining images with useful information, LPR locate the plate from the image, find the location of the plate, and automatically segment the license plate to get the image of characters, then identify the characters one by one, finally the recognition result are output to the display device.For LPR, the first step is to detect vehicle, that is to determine whether there is vehicle coming, detection methods are generally divided into two types: hardware-based trigger and video-based detection. Hardware-based license plate recognition method is triggered by a trigger to determine whether a vehicle passes by, some flip-flops can achieve the velocity measurement and direct license plate recognition, and video-based detection of license plate recognition is only to use necessary equipment including cameras, video capture cards and computer, the rest of the work entirely base on a computer program, namely, the license plate recognition algorithm routine to realize.License plate recognition method mainly includes three parts, namely, vehicle license plate location, vehicle license plate character segmentation and character recognition, only the combination of the three parts of the organic to get a satisfactory recognition output.License plate location is the most important part in the whole system, the success of the license plate location is a prerequisite for the follow-up Character Segmentation and Recognition, if you can not locate a license plate or the license plate image located is not ideal, system will be failure.License plate character segmentation is also very important, its role is to facilitate with the license plate character recognition, obviously, recognition of image with many characters is more difficult than recognition of a single character, therefore it is quite necessary to cut down the characters before they are identified. License plate character recognition is also important, cause it is directly related to the system's recognition results. License plate recognition system is the ultimate goal of the content on the license plate, namely, to identify characters. Some license plate recognition system is used in the method of license plate characters in undivided, take character recognition as a kind of pattern recognition processing which is about a string of characters. Therefore often, character segmentation and recognition are classified into a category, collectively referred to as character recognition.License Plate Recognition process need lot of Knowledge of image processing, Computer vision, Computational Intelligence, morphology, and so on. License Plate Recognition process need lot of Knowledge of image processing, Computer vision, Computational Intelligence, morphology, and so are. ALL license plate identification will be specific to a certain frame finally, for the image features to carry out image processing is also the steps which is adopt by all license plate recognition system. Both digital photo collected by camera or video frame captured by the camera, must contain a lot of noise or interference information which can only through the grayscaling, arization, image enhancement techniques and so on, to be filtered out, each method of image processing techniques are not fixed, for different situations we need to make the correct or practical choice. Since it is a license plate recognition, then the object is to identify the license plates, a variety of inherent characteristics of the license plates is the necessary condition for each recognition algorithms. License plate features include geometric features, color gray-scale characteristics, texture characteristics and the characteristics of the contents of its internal character, etc. For different characteristics of different recognition algorithms have been proposed, but these algorithms are not merely based on a characteristic, often a combination of several features.Most of the license plate identification license plate recognition system is based on a static image processing, one or a few frames are captured by camera or video camera for license plate segmentation and recognition operations, Presented in this paper based on multi-frame images and dynamic video license plate recognition technology, can locate the license plate which is captured by video camera many times for moving vehicle in the video and find out which is relatively better to carry out follow-up operation of license plate characters and Segmentation.Presented in this paper is a kind of video-based detection license plate recognition method. The entire license plate recognition system is divided into three modules, namely the vehicle video detection, location and license plate character recognition(including License Plate Segmentation). Each module can be a separate simulation experiment, basically used realize the relevant technologies and functions.Video detection part, the authors improved the relatively poor real-time background subtraction method, using adaptive background amendments to get license plate information that contains the key frame.In License plate location part, this paper, for the license plate location of every single frame image, first rough positioning, then precise positioning, use a kind of license plate method which is based on regional shape and gray-jump, Thus, to get a license plate template, through vehicle tracking technology which is use least squares technique to estimate the location, use a kind of template matching technique which is based on edge lines and plate texture features to update license plate template, When you reach certain conditions, namely the successful completion of the positioning plates can be used to determine the character segmentation and recognition operation of license plate image.In the license plate character segmentation and recognition part, according to the unique geometry features, gray features, edge features, character content features, etc. this paper, use the improved gray-scale approach, through Training and matching the character sample which is segmented out, realize the output of the character.This paper is based on platform established by C++ language, through making use of Open computer vision library (Open CV) and DirectShow which is a kind of stream processing tools, etc. to experiment. Compared with other method, in certain conditions, the license plate recognition system in this paper could meet the real-time and accuracy requirements well.