本論文主要是設計一個低成本、高實用性的銀行保全監控系統。當客戶進入銀行等金融機構時，此系統會即時識別客戶是否有遮蔽臉部之裝扮(例如頭戴安全帽或口罩)，若有此裝扮時，系統會立即啟動語音予以告知客戶將此裝扮褪去，同時亦可提醒保全人員留意其犯案之可能性。 系統架構於個人電腦，並透過USB介面的數位攝影機作為影像擷取裝置，可降低成本並提高實用性。利用影像處理技巧和類神經網路等技術，完成影像辨識，本系統可在複雜背景環境下執行，亦可同時偵測多目標，在辨識率方面，每一類樣本的辨識成功率均可達到九成以上。最終目的在協助保全人員，避免人為疏忽而使得歹徒有可乘之機。 This thesis is mainly designed for a low-cost and high practical monitoring system on bank security service. As any customer entering a financial institution, e.g. a bank, this system will recognize if the customer has any decoration to cover his/her face. If it is true, the system will launch its sounds to inform the customer of removing the decoration immediately. Meanwhile, the system can remind the security service personnel of noticing the possible criminal. This system is built up in a personal computer. A digital camera with USB interface is used for the image collector; it can reduce the cost and raise the practicability. The techniques of image processing and neural networks are used to finish image recognition. This system can execute in a complex background, and detect many targets. The recognition rate will be more than 90% in each type of samples. The final goal of this thesis is to help security service personnel with avoiding man-made carelessness of robbery.
