Discussion on license plate recognition technology

  • Detail

Discussion on license plate recognition technology of traffic technology

license plate is the only mark for vehicle identification in the world. Although the characters, colors, formats, contents and production materials of the license plate will be diverse, the license plate is still the most accurate and specific identification mark in the world

according to relevant statistics of international transportation technology, 78 companies worldwide, except China, are producing license plate recognition products

technical indicators for evaluating the license plate recognition system

technically evaluating a license plate recognition system has three indicators, namely, recognition rate, recognition speed and background management system. Of course, the premise is that the system should be able to operate stably and reliably

I. recognition rate

the most important indicator of whether a license plate recognition system is practical is the recognition rate. International transportation technology has made a special discussion on the recognition rate index, which requires that the correct recognition rate of 24-hour, all-weather and all-weather license plates is 85% - 95%. The license plate recognition system of lip vision has achieved a correct recognition rate of more than 90% of the whole license plate in practical application

in order to test the recognition rate of a license plate recognition system, it is necessary to install the system in a practical application environment, operate for more than 24 hours a day, collect the license plates of at least 1000 vehicles passing in natural traffic flow for recognition, and store the license plate image and recognition results for retrieval and viewing. Then, we also need to get the actual vehicle image and the correct manual recognition results. Then the following recognition rates can be counted:

1. The recognition rate of natural traffic flow = the total number of correctly recognized full license plates/the total number of vehicles actually passed

2. The percentage of recognizable license plates = the total number of correctly read license plates manually/the total number of vehicles actually passed

3 The correct recognition rate of the whole license plate that can be recognized = the total number of correctly recognized license plates of the whole license plate/the total number of manually read license plates. These three indicators determine the recognition rate of the license plate recognition system, such as reliability and error recognition rate, which are the intermediate results of the license plate recognition process

second, recognition speed

recognition speed determines whether a license plate recognition system can meet the requirements of real-time practical applications. A system with high recognition rate, if it takes a few seconds, even a few minutes for some wear-resistant, heat-resistant, food contact parts such as coffee machine and juice machine, to recognize the results, then this system will be of no practical significance because it cannot meet the real-time requirements in practical applications. For example, in highway toll collection, one of the functions of license plate recognition application is to reduce traffic time, and speed is a powerful guarantee to reduce traffic time and avoid Lane congestion in this kind of application

the recognition speed proposed by international transportation technology is less than 1 second, the faster the better. The license plate recognition system based on Lipp vision has an average recognition speed of 200 milliseconds in practical applications

III. background management system

the background management system of a license plate recognition system determines whether the license plate recognition system is easy to use. It must be clearly recognized that it is impossible to achieve 100% recognition rate, because the license plate is dirty, blurred, blocked, or the weather may be bad (snow, hail, fog, etc.). The functions of the background management system should include:

1. The reliable storage of recognition results and vehicle image data. When the multi-functional system operation makes the network error, it can protect the image data from loss, and it is convenient for manual troubleshooting afterwards

2. Effective automatic comparison and query technology. The recognized vehicle license plate number should be compared with thousands in the database ● exchange the license plate number of the servo speed regulation system for automatic comparison and prompt alarm. If the vehicle license plate number is not read correctly, fuzzy query technology should be used to obtain the relatively "best" comparison result

3. For joint operation, a good license plate recognition system also needs to provide real-time communication, network security, remote maintenance, dynamic data interaction, automatic database update, hardware parameter setting, system fault diagnosis; Lipp vision's background management system adopts a multi task parallel processing mechanism to integrate the front-end license plate automatic recognition and the back-end image with the help of this transaction image database management, which reliably ensures the storage and management of image data and recognition results. At the same time, the front-end can operate background data query, statistics, printing, storage and communication in parallel, without affecting each other. The background operation data query, statistics, printing, storage and communication will not affect the front-end real-time vehicle capture and license plate recognition. Lipp vision version of the license plate recognition system software, installed on the system client and server workstation, can build a fully functional vehicle deployment network. The operator remotely manages and maintains the system of each client workstation through the monitoring and management program at the central server workstation, including automatic data upload and distribution, setting operation parameters, viewing system operation status and abnormal information, etc

choice from practical application

even a license plate recognition system that meets the practical standard, due to the different technical route, software and hardware architecture and trigger mode, its effective function depends on the actual application requirements

I. technical route of license plate recognition

the process of using computer vision technology to recognize license plates usually includes five steps: vehicle image acquisition, license plate location, character segmentation, optical character recognition, and output recognition results. The acquisition method of vehicle image determines the technical route of license plate recognition. At present, the two mainstream technical routes of international its are natural light and infrared light image acquisition and recognition. Natural light and infrared light will not have adverse psychological effects on the human body, nor will they produce new electronic pollution to the environment. They belong to green environmental protection technology

natural light route refers to the use of natural light during the day, the use of auxiliary lighting sources at night, the use of color cameras to collect true color images of vehicles, and the use of color image analysis and processing methods to identify license plates in Jinan assay low-temperature tank. The route of natural light true color recognition technology is consistent with the photosensitive habits of human eyes, and the true color image can reflect the real image information of the vehicle and its surrounding environment. It can not only be used to identify the license plate, but also be used to identify the color of the license plate, traffic flow, vehicle type, vehicle color and other vehicle characteristics. The images collected by one camera can realize the acquisition, recognition of all front-end basic video information and manual image forensics and discrimination at the same time, which can reserve interfaces for future intelligent transportation system projects. Lipp visual selection from

Copyright © 2011 JIN SHI