| Summary: | 碩士 === 國立暨南國際大學 === 電機工程學系 === 95 === This thesis designs a real time ATM security system which could avoid the bandit draws money from ATM. When a person uses the ATM system, the security system will be triggered and identify whether the user’s face is covered or not. If the system finds the user’s face is covered, the ATM is locked and the alarm is working, until the user takes off the covers. By the way, the security system also records the image of the user’s face.
Based on the skin color quantity and the human’s face features, the security system has three stages to identify whether the user’s face is covered or not. If the pre-stage is passed, then the ATM user will go to the next stage. If all stages are passed, the user’s face is identified to be not covered and he can draw money from the ATM. Stage 1 extracts the face area from the image and checks the rate of the skin color in the face area. If the rate is larger than the threshold, then this stage is passed. Stage 2 uses connected component label algorithm to calculate the rate between the width and length of the skin color. By using the normal rate of the face, we can avoid the mistake of skin-color area which is not face. Stage 3 uses the feature of human face to identify whether the human nose and mouth are covered or not. The experiment result shows that the proposed security system works successfully.
|
|---|
