本論文提出使用兩階段的辨識方法應用在寵物狗的身分識別，針對輸入影像以其外在的生物特徵進行初步的分群，將外在特徵相似的影像劃分到同一群集，並以臉部定位對影像進行校正，以消除拍攝照片時因外在因素所產生之誤差並將影像以狗臉的邊框進行正規化，再以LBP進行特徵轉換獲得紋理的特徵圖，並以孿生網路的架構對正規化後的影像進行影像比對，以歐式距離計算輸入影像與系統資料庫已註冊影像相似度，我們以農委會提供的犬隻資料集進行實驗，可達87%的辨識率。;In this paper, a two-stage identification method is proposed for pet dog identification. The input images are initially grouped by their external biometric features, and the images with similar external features are classified into the same group, and the images are corrected by face feature localization to eliminate the errors caused by external factors when taking photos.The similarity between the input images and the registered images in the system database was calculated by using the Euclidean distance, and we were able to achieve 87% recognition rate by using the data set of dogs provided by the Council of Agriculture Executive Yuan.
