Anomalous-Aggressive Event Detection Techniques

Human behavior refers to the way humans interact with their surroundings. It can be classified as either being normal or anomalous behavior. Anomalous events are behaviors that are deviant against all societal norms. One event makes another event happen, where the cause-effect relationship in behavior is the law of science. Automatically detecting anomalous events has become an important research area. It plays a significant role in video surveillance security, especially in public places also human–computer interaction. Recently, researchers have oscillated between traditional learning, deep learning, and hybrid-based models. The traditional learning-based model depends on extracting the best features to contribute to the result. The deep learning-based model deals directly with data and skips the manual step of feature extraction, offering an end-to-end problem solution. The hybrid-based model combines the benefits of both models to improve the detection result. This paper presents an inclusive review of traditional learning, deep learning, and hybrid-based models for anomalous event detection. It also hands out discussions on state-of-the-art approaches used.