| Summary: | 碩士 === 國立聯合大學 === 機械工程學系碩士班 === 103 === Nowadays, Cars and motorcycles are necessary transportation for the public. Recently, many kinds of automatic systems for stolen-car investigation, ETC debit and smart car-parking management have been derived from license plates automatic recognizing systems under the rapid development of research environments.
The multiple specifications of license plates recognizing system is programmed with the built-in functions of Matlab software for the research. The samples for the research are the vehicles taken photos from those parked besides roads with different distances of 100, 180, 260 centimeters to verify the better distance for recognizing. The different surroundings of parked cars are also tested to study the influences to the system.
There are three major steps in our research on license plates recognizing system, including License Plate Positioning, License plate segmentation and License Plate Recognition. The methods in the research including grayscale, edge detection, image enhancement and median filter have been used to facilitate the recognizing system. At first, the positions of license plates are searched vertically and horizontally. Secondly, the plates with matched characters will be segmented by built-in mask. Finally, the recognizing result will be gained by template matching.
There are total 671 license plates that have been tested by the system as samples. The results reveal the success rate of recognizing is 98.1% as a whole. The success rate on license plates positioning is 96.8%, 96.4% for old-style plates and 97.4% for new-style ones. The success rate for integral characters recognizing is 91.8%, 89.2% for segmentation of old-style plates and 93.5% for new-style ones.
|
|---|
