| Summary: | 碩士 === 國立臺北科技大學 === 資訊工程系研究所 === 101 === This study presents a classification-based License Plate Recognition (LPR) for the unclear and degraded License Plates (LPs). The proposed system adopts Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM) as the feature descriptor and classifier to implement a two-level LPR, including LP localization and character recognition. For the recognition efficiency of the proposed system, two trained databases for LP localization and character recognition are reduced by the proposed mathematical and feature-based approaches, and thus the computational costs on searching process can be significantly saved. Moreover, to effectively and adaptively operate the proposed system in real traffic monitoring applications of the proposed system, an online updating mechanism is proposed and implemented. Besides, to keep the quality of the training databases, the online updating mechanism verifies LP samples based on their own image features. To show the experimental results, the proposed system is demonstrated by real unclear and severely degraded LP images. With the online updating mechanism, the proposed system performs feasible and effective recognition results for the Unclear and Severely Degraded License Plate Recognition (USDLPR).
|
|---|
