邊界切割方法在車牌辨識系統之應用

Abstract

在本篇論文中，提出兩個新的應用方法來解決車頭燈及散熱器還有光線變化跟複雜的車子圖騰對車牌辨識系統所造成的干擾。在此論文的第二章中，提出一個名為”雙邊界判斷法”(double edge method)來處理車頭燈跟散熱器所造成的干擾，導致係系統誤判。此原理是採用車牌與車頭燈和散熱器不同的特殊特徵來判斷候選區域中使否含有車牌。而這一個方法可以將車牌擷取的成功率提升到98.4%。除此之外，在第四章中會介紹邊界切割方法(edge segmentation method)，這一個方法是用來處理光線變化跟複雜圖騰所造成的切割字元方面的困難，而且成功的將字元切割成功率提升到99.67%。

This thesis proposes two methods, the double edge method and the edge segmentation method, to enhance the accuracy of the LPR system. After the process of projection extraction method, some potential license plate areas can be found in an image, which may includes the headlamps and radiators. The double edge method is then utilized to determine the precise area of the license plate, not the areas of the headlamps or radiators. With the double edge method, the accuracy of the license plate extraction is highly increased up to 98.4%. As for the edge segmentation method, it is proposed to extract the characters from the license plate area just obtained and this area is often badly influenced by illumination variation and complex texture of a vehicle. With the edge segmentation method, the character extraction rate can be increased up to 99.67%. Finally, experimental results have been also included to demonstrate the success of these two proposed methods.