Gradient -based license plate recognition algorithm template feature
|School||Huazhong University of Science and Technology|
|Keywords||License Plate Location License plate character recognition Gradient Template Features|
With the sustained socio-economic development, China's number of motor vehicles at a higher growth rate. However, the field of transport infrastructure is still far behind the rapid growth in the number of vehicles , which caused a series of social problems , traffic congestion has become a common phenomenon in cities . In order to improve the utilization of transport resources , improve transportation efficiency , ease traffic congestion, reduce energy consumption , intelligent transportation systems (Intelligent Transportation Systems, ITS) become a national hot research issue , has been widely appreciated , developing very rapidly. Vehicle license plate recognition (license plate recognition, LPR) as the intelligent transportation is an important concept is to achieve traffic monitoring and management automation major technologies. And its access to information as an important means for traffic monitoring and management system provide important data. Vehicle license plate recognition generally include location, license plate character recognition and image processing related to three main parts . In license plate location study, proposed the use of gradient template features precise positioning plate segmentation algorithm with the template . The algorithm is focused on the characteristics of the overall structure of the plate area , high positioning accuracy can be obtained . Meanwhile, the analysis of the Cascade cascade classifier algorithm in license plate positioning tasks in the application, and thus constitutes a high positioning accuracy and efficiency of the license plate location classifier . In the license plate character recognition studies, using pixel-based connected domain growth plates tilt correction algorithm for license plate image correction . Using statistical characteristics and character-based joint characteristic geometric features to identify the characters on the license plate . The experiments show that the proposed license plate location algorithm has higher accuracy and reliability , the success rate was 95.4% coarse positioning , precise positioning success rate was 92.3% , positioning speed of each image 300 ~ 500ms. And the contrast is low, the license plate is stained , etc. have a higher robustness. In the license plate recognition experiment, license plate character recognition algorithm recognition rate of 93.59 percent overall , the character recognition rate 98.47% , reached a relatively satisfactory level.