大多數的汽車自動辨識系統都是使用車牌辨識來作為第一線的汽車識別，車牌辨識系統雖然能讀取大多數的車牌，但是可能會失敗於兩種情況，一種是當車輛經過時，無法辨識出高信心值的車牌號碼，另一種則是車牌辨識系統辨識錯誤，造成車牌資料錯誤。 本論文利用SIFT特徵辨識的方法，用來處理被車牌辨識系統拒絕的車輛，當現行車牌辨識系統在面對車牌遮蔽或塗汙等影響車牌號碼辨識時，在第一次出現時先經由人工判定車號，再將人工判定車號及建立對應的特徵進行配對，透過車牌特徵辨識系統之技術，針對這些因車牌遮蔽、塗汙等特定車輛進行處理，之後由此特徵比對技術輔助辨識正確車號，並可建立自動車號特徵比對作業，減少事後人工補登及辨識引擎誤判車號之數量，降低相關檢核作業之人力成本。 ;Most of automatic vehicle recognition systems use an ALPR as the first line-of-processing for recognizing vehicles. The ALPR reads the majority of the plates. However, the ALPR may fail in 2 cases: 1) the ALPR cannot recognize the plate due to the masking and smearing, and 2) the ALPR misrecognizes the plate. In this study, we propose a license plate authentication system based on SIFT algorithm to process that cannot be recognized from the ALPR. When a car with a masking or smearing plate first appears in the ALPR, the license plate number is determined manually. Then, the SIFT algorithm is applied on the image of vehicle with masking and smearing license plate for extracting features. Finally, while a vehicle cannot be recognized from the ALPR appears, it would be authenticated through the SIFT feature matching. Based on the proposed scheme, the manpower costs associated with the inspection operations could be reduced magnificently.
