基於霍夫轉換的側面影像口罩偵測技術

The Gauze Mask Detection Technique for Silhouette Image Based on Hough Transform

10.29943/DNSJ.201006.0085

陳一斌；洪維偉；劉力魁

Face detection ； canny ； Hough Transform

電腦視覺、影像處理與資訊技術研討會

2010（2010 / 06 / 01）

85 - 90

繁體中文

現今的人臉辨識防盜系統主要是以正面人臉特徵來做為辨識依據。然而，以目前的技術，對於遮蔽的或偽造的人臉等問題，仍然有困難與死角無法完全解決。本系統是針對銀行提款機室的防盜安全，以及對舊有系統的不完整，提出解決的辦法。我們使用基於霍夫轉換的方法，通過偵測側面影像是否有可疑的直線來判斷目標有沒有戴口罩，首先我們由安裝在側面的攝影機取得要偵測的影像，先將影像做灰階化處理，接著使用canny找出影像的邊緣，再經由頭部辨識來找出口罩的可能位置，最後再以Hough Transform偵測出可疑的直線來進行判斷。

Face detection based burglarproof systems used today are mainly depended on the features of front face. But some problems like the fake face, covered face, etc., still can't be solved by the technologies used today. Here we propose a method to solve the problem of the traditional ATM burglarproof system. This Hough transform based method uses the silhouette image to determine whether the face covered with gauze mask by detecting line features. First we get the silhouette image. Then change the image to be a grey image, and use the Canny operation to find out the edge of object. After this we detect the head position, and further determine the possible positions of the gauze mask. Finally the Hough transform is used to detect the gauze mask.