透過您的圖書館登入
IP:18.223.0.53
  • 學位論文

類神經網路及影像處理於車輛特徵辨識之應用

Vehicle Feature Recognition Using Neural Networks and Image Processing Techniques

指導教授 : 李錫捷
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


現今關於車牌辨識系統的研究已非常的多,但對於車輛顏色辨識與車型識別的研究卻是非常少見,如果能得知這些車輛特徵的資訊,警方在追蹤可疑車輛時,必定能夠有效的協助警方節省過濾可疑車輛的時間,因此車輛特徵的確是一個非常有用的資訊。如果拍攝的影像是在室外,環境因素會較為複雜,如曝光、陰影、污損、變色或裝飾品等等,都會造成車輛特徵辨識上的困難,因此我們必須克服環境複雜的問題。 本研究提出了具有適應環境複雜能力的車輛顏色的辨識與車型的識別。車輛顏色辨識系統主要分為擷取感興趣的顏色區域、特徵值的正規化、顏色辨識等三大部分。本研究使用了975張樣本進行實驗,實驗結果在車輛前車牌顏色辨識率平均達91.18%,在車輛後車牌顏色辨識率平均達91.17%,整體車輛顏色辨識率平均可達91.18%,而車型識別率平均則達95.34%。

並列摘要


In recent decades, automatic license plate recognition has been widely studied. However, there are relatively few studies focus on the vehicle color recognition and vehicle type classification. combined with three vehicle features, license plate, color, and type, it would be easier for police to identify a specific vehicle. It can help police in saving time to filter the suspicious vehicle effectively. However, it is usually difficult in getting correct vehicle features since the vehicle image taken could be overexposed, underexposed, shadowed, or color fading and the vehicle itself could be damaged or decorated. This study aims to propose a system that has the ability of vehicle color recognition and vehicle type classification to adapt itself to the complex environment. The vehicle feature recognition system may be divided into three major components: ROI (Region of Interest) detection, feature extraction and normalization, and color recognition. To demonstrate the performance of the system, there are 975 different images used for conducting the experiments. The results of this study have the average accuracy of 91.18% for color recognition in the case where images are captured in front of the vehicle and 91.17% for color recognition in the case where images are captured behind the vehicle. In addition, the average accuracy is 95.34% for the vehicle type classification.

參考文獻


[4] 蔡孟達,「色彩影像初階校正與縮放技術研究」,私立中原大學,資訊工程研究所,碩士論文,民國94年。
[6] 陳志凱,「應用類神經模糊系統於影像顏色辨識之研究」,私立大同大學,電機工程研究所,碩士論文,民國94 年。
[3] 王振興,「多標的汽機車車牌辨識系統之研究」,私立元智大學,資訊管理研究所,碩士論文,民國92年。
[1] 仲崇實,「影像處理與類神經乏晰方法於車輛牌照自動辨識之應用研究」, 私立元智大學,資訊管理研究所,碩士論文,民國87年。
[2] 魏銪志,「動態多標的車牌辨識系統之研究」,私立元智大學,資訊管理研究所,碩士論文,民國89年。

延伸閱讀