Algorithm And Implementation Of Automatic Vehicle Identification System

With the economic development and rapid increase in the vehicle amount in our country, road transport has become the important form of transportation and a focused infrastructure. Constantly blocked city transportation needs more advanced and more effective traffic management and control. The Intelligent Transportation System (ITS), which improves efficiency and safety using electronic information technology, has become the main direction for transport administration.By means of monitoring the road traffic flow information in real-time, the ITS is capable of performing various traffic control and management tasks, making full use of related transport facilities. The License Plate Recognition (LPR) system is the key component of ITS; it can accomplish diverse functionality, including automated vehicle supervision, verification, registration and alarming. Due to this, LPR is widely used in highway administration, parking-lot charging, neighborhood vehicle management and electronic policing.This thesis presents the algorithm and software design for civilian license plate recognition. The main work is composed of:(1) License plate location algorithm. The source image is divided into several rectangle subimages and the locate mask is obtained by analyzing every subimage locally in the color dimension. Then the mask is de-noised by means of mathematical morphological processing. After some license plate candidate is extracted from the locate mask, Hough transform is performed to correct the acclivitous license plate. Then every candidate is checked against the size and texture features of license plates and the false candidates are removed. License plate is located accurately and the unacclivitous license plate is forwarded to the next stage.(2) Character segmentation algorithm. Each rectangle region containing a connected component in the binary license plate image is extracted based on the region-growing approach. The character size and duty-ratio features are utilized to determine whether or not some region is a reasonable candidate. Then the character spacing characteristics are used to eliminate the regions whose position is inappropriate and supplement the regions which are missing.(3) LPR software system. A C-language-based LPR demo system is built from scratch. The algorithms designed in this thesis are incorporated into this system. This software is available to the public, easy to change or add features, and hardware-independent.The algorithms proposed in this thesis have been implemented as an executable program. A large amount of representative vehicle images taken from diverse conditions are used as testing source to verify our algorithms. The results illustrate that the success rate of our location algorithm is 96.3% while the success rate of our segmentation algorithm 96.8%.