建立于android系统之钥匙自动辨识系统-台中区网中心.pdfVIP

TANET2013臺灣網際網路研討會- 【論文集】 建立於 Android 系統之鑰匙自動辨識系統 王建仁 沈英謀 蔡哲民 鄭承淵 崑山科技大 電腦與通訊系、資訊管理系、資訊傳播系、電腦與通訊所 { cjw, shenyin , tjm }@.tw ; jim122664@livemail.tw 摘要 select the one in the database which is the most similar to the testing key as the recognition result. The features used in our system are shape and txture. In 生活中有許多地方都必須利用鎖來防止財物 order to avoid the influence resulted in different 被他人竊取，所以一般人都會擁有一把以上的鑰匙 position and angle of the key, after segmenting the key 來解開各種不同的鎖具。由於鑰匙在外型上都很相 image from the background, the system would first 似，對於視力模糊、記憶力退化的年長者或是其他 rotate the key and align the key image on the y-axis of 具有類似障礙的族群來說，可能難以分辨出鑰匙的 the test plane. Then the key image is scaled to a 種類。 predefined size. The feature of the shape is the 本論文提出一套Android Based的鑰匙自動辨 Centroid-contour-distance (CCD). The texture feature extraction is as follows: The key image is 識系統，使用者只須使用智慧型手機的 Camera 拍 processed by Sobel edge-detection algorithm first, and 攝欲辨識的鑰匙，系統就會自動從影像中切割出物 then the divided into N regions and the edge points of 件，與特徵資料庫進行比對，並將分析出來的鑰匙 each region is counted. These edge point counts 種類回傳給使用者。特徵資料庫主要分為「形狀」 serve as the texture distribution. The K-nearest 與「紋理」兩種特徵，為了防止使用者物件擺放的 neighbor (KNN) method is applied to perform the 角度與位置的不同影響辨識結果，系統會先進