Abnormal Face Detection Technology And System Development For Intelligent Security Of ATM

In recent years, with the construction of various electronic business of bankdeveloping rapidly, the number of ATM increase rapidly, and usage also spread to thecountry. Many criminals start to make damage to ATMs because of ATM with thefeature of automatically saving money, depositing money and sending money and theunattended environment. On the one hand ATM crime makes the majority ofcardholders suffering a huge economic loss, on the other hand it also disrupt thenormal working order of financial institutions, banks and other financial institutions,bring a certain loss. From the numerous ATM crimes, in order to evade surveillancecameras on the ATM surveillance, many suspects in the crime are in appearancedisguise, for example wearing sunglasses or wearing masks, obscuring the faceinformation. This behavior not only increases the withdrawals’ crime suspect, butalso brings difficulties to the policeman to catch the criminal suspect. Therefore,detecting the disguised face of customer is effective to catch withdrawals’ crimesuspect.This paper proposes a method of detecting disguised face (sunglasses and masks)for ATM surveillance video of the front face basing on the requirements of bank, fullyconsidering the characteristics and scene changes of ATM surveillance video of thefront face, with some accuracy, timeliness and robustness. To solve the disguised facedetection problem, this paper collects a large number of ATM positive video,analyzes the scene feature, and designs an algorithm of processing the specificdetection. The detection algorithm can detect human in the video and locate faceusing background subtraction and AdaBoost with Haar-like feature and HOG feature,proposing a sunglasses detection method basing color and a mask detection methodbasing line detection. At last, we develop an abnormal face detection prototypesystem.