Publication

Abstract : Automated Teller Machines (ATM) is an essential and inevitable part of our daily life, the money transactions become safe and fast with the help of ATMs, still, the machines and the ATM booths are prone to the severe criminal activities. In any case of criminal activity, the victims will be helpless when the booths are located in remote areas. According to the current scenario, the investigation officials will be aware of the crime activity only after the crime is over and they will try to track the criminals using the recorded data in surveillance cameras. The proposed study is aimed to identify the abnormal activity in ATM Booths with the help of artificial intelligent techniques. The Support Vector Machine (SVM) strategy is employed to classify the activity into normal and abnormal class, the continuous video streams are encoded by the histogram of gradients technique, and the feature mapping is made by K-means clustering algorithm. The parameters of the SVM classifier algorithm is tuned by the proposed hybrid HGOA optimization algorithm. The performance of the proposed model is evaluated for human motion identification and the anomaly event is identified on two benchmark datasets such as HMDB-51 and CAVIAR datasets.