Face identification in video is a difficult and fascinating topic, especially when it's used in Automated Teller Machines (ATM). Covering one's face with items such as a mask, scarf, or sunglasses is a popular illegal conduct in ATM robbery. As a result, reducing robberies and other crimes can be accomplished by deploying ATM security cameras to detect covered faces. In this paper, a new method is presented for detecting covered faces in ATM surveillance system. To achieve precise foreground, a new and expedient foreground extraction method is recommended. In this research work, we propose binarization with Otsu’s Thresholding method to find the intensity value of pixel that subtracts background and foreground from a given input image. To determine whether the human face is covered or not, we then locate the face using the Histogram of Oriented Gradient (HOG) approach.The results of our experiments demonstrate that our algorithm achieves a high detection rate while maintaining a low false negative rate.

Suspiciousbehavior,Face detection, Otsu’s, HOG, Video Surveillance