ANOMALY HUNTER: YOLOV3 ADVANCED DEEP LEARNING MODEL FOR HUMAN ABNORMAL DETECTION
P. Deepan, Dr. B. Rajalingam, Dr. R. Santhosh kumar, Mr. N. Arul · 2024
The identification of aberrant human behaviour is an important component in the process of protecting the health and safety of persons as well as the preservation of the safe environment in public areas. Advanced algorithms such as YOLO (You Only Look Once) and Convolutional Neural Networks (CNN) are utilised by emerging technologies such as human behaviour and anomaly detection. These algorithms are utilised to extract characteristics, manage temporal relationships, and enhance the accuracy and efficiency of human behaviour detection systems. Examples of such algorithms are the YOLO and CNN algorithms. This approach was designed with the purpose of being utilised in real-world contexts, and more specifically for recognising suspicious behaviours in surveillance video collected by closed circuit television (CCTV) cameras. The technique was developed with the intention of helping to identify suspicious actions. This model takes use of the object identification approach known as YOLOv3 in order to recognise human actions and abnormalities contained in the video data. Additionally, a Convolutional Neural Network (CNN) is utilised in order to extract action characteristics from each tracked trajectory. This is done after the previous step has been completed. In the end, but certainly not least, a You Only Look Once (YOLOv3) is utilised in order to construct a model for the purpose of recognising abnormal behaviours. This model enables the prediction of abnormal actions that are carried out by individuals. To summarise, the functioning of this technology entails getting video information as input from security cameras and then employing a variety of complicated algorithms in order to identify and categorise anomalous actions. This is done in order to prevent and detect any potential threats. By utilising YOLO and CNN for the purpose of feature extraction and anomaly identification, the system is able to accurately recognise and anticipate abnormal types of human activity. This is made possible by the usage of these two techniques.