Font Size: a A A

Multi-view Research On Obscured Objects Of Traffic Image

Posted on:2015-05-29Degree:MasterType:Thesis
Country:ChinaCandidate:L Y HuiFull Text:PDF
GTID:2298330452958904Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of social economy, the number of domestic carsincreases highly recently. As the only certification of car identity, the license platenumber is widely used in vehicle detection, identification and tracking. With thehardware enhancement of intelligent traffic system, it’s possible for multi-viewmonitoring which will achieve a highly attention in the future. Plate recognitionalgorithm and multi-view fusion algorithm which is based on obscured traffic imageis studied in this article, the major works are as follows:1. The pre-processing technology of image in plate recognition system, includingcolor image gray processing, adaptive binarization, Gauss filtering, canny edgedetection, etc, which is better for license plate recognition.2. Extraction of the license plate is affected directly when occlusion occursbetween vehicles which may lead to inaccurate and false identification. In the processof recognition, based on the rapidity and accuracy, improving plate recognitionalgorithm which is currently used. Firstly: according to the principle which is fromcoarse to fine, achieving the positioning of plate fast. Using corner detection to locateplate area in the collected images, then makes a fine processing to determine accurateposition of plate; Secondly: achieving exactly recognition of character through theoryof image enhancement. Enhancing the positioning plate and filtering out irrelevantcharacteristics with character, the character can accurately identified when matchedwith template.3. Adopting multi-view fusion technology to make integrated decision for licenseplate. Two views are adopted to capture images and each view is recognizedseparately in this article, then each recognition result is fused through confidencepolicy, finally the license plate number is determined.This article adopts Visual C++6.0to make programming. Experimental resultsshow that the proposed theory and method has higher accuracy and computationalefficiency. It has some theoretical innovation and social use values.
Keywords/Search Tags:intelligent traffic, plate recognition, multi-view fusion, corner detection, confidence
PDF Full Text Request
Related items