Research And Design Vehicle Recognition System Base On Vehicle Surface Image

Vehicle recognition is one of the important branch of Intelligence Traffic System,and the main application field of ITS.Vehicle recognition plays an important role in the charge management system of expressway.Due to the vehicle face regional have abundant Vehicles' characteristic information,it has great significance for research on vehicle recognition technology based on vehicle surface image.The paper researched and designed the Image Data Acquisition Platform based on USB interface.The image data are stored into SRAM by CPLD and Imaging Sensor's sequential coordination on this platform,and then this image data are read by DSP,finally this image data are transferred to the host computer by HPI interface of DSP and general programmable interface of USB's communication.This paper mainly studies vehicle recognition technology based on image processing and pattern recognition,and a practical vehicle recognition system was designed by C++ on host computer,specific contents are as follows:Firstly,the paper used gathered vehicle surface image to do feature location and extraction.Vehicle face regional was observed and analyzed can be drawn the main feature location are vehicle-logo and headlamp.At first,vehicle face regional can be located by horizontal integral projecting and edge detection operator,then vehicle-logo can be located by horizontal and vertical integral projecting,and headlamp can be located by HOUNG transform and convex hull algorithm and Snake model,next used HU invariant moment to extract eigenvalue.Finally,eigenvalue was matched by Euclidean Distance.Analysis of recognition results this recognition algorithm achieves good recognition effect.Then,according to the fact that license plate location algorithm on the problem of location performance is not ideal on vehicle recognition,the gray stretch and maximum variance thresholding method of image pre-processing was proposed,the selection on the structure and color of license feature and verification conditions were used to identify whether the position areas are license plates,then remove the false region.This method improved image contrast and automatic selecting for binary threshold.Experimental result shows that this method has high accuracy and great rapidity on the license plate location.Finally,used C++ programming language and great frame design of MFC,and combined with background data support of MySQL database.Consideration from two aspects of real-time and practicality,vehicle recognition system development and designed based on vehicle surface image.This recognition system can detect and identify and classify the vehicle automatically,which can be used as the basis for traffic charges,management,scheduling,statistics and inspection.Experimental result shows that this recognition system achieved the expected design effect and has a good practicability.