| Summary: | 碩士 === 元智大學 === 電機工程學系 === 99 === The management and security for automobiles has become much more important due to the increasing population of automobiles for a modern society. Automobiles license plate recognition system will provide an efficient way for automatic surveillance in tracking stolen cars and any illegal actions such as copying license plates. Besides, some environmental factors such as weak lightings and abnormal weathers would increase not only detection difficulties but also rises false recognition results. These are many problems we need to overcome to attain high recognition rates.
This thesis mainly proposes and designs an image processing procedure for license plate recognition. This procedure includes image preprocessing and gradient analysis combined with parallel computing in GPU structure for license plate, tilted license plate correction, the license plate region framing, license plate normalizing and interfered noise filtering. In addition, this thesis exploits horizontal and vertical projection method to get the height and width pixels of strings and to cut the characters to be recognized. Segmented characters require normalization. These will become standard images to be compared with information of incoming images. The tasks involved will include template matching, feature matching and simple characters classification etc.
License plate recognition system potentially could be combined and coordinated with a related data bank of license plates to effectively perform a large scale plates recognition task. Furthermore, this system can be applied in automobile inspection, automobile management for buildings and automobile intrusion detection etc.
|
|---|
