Review on Anomaly Detection Using Machine Learning and Deep Learning Techniques

Object detection is a critical task that focuses on determining the position of a target in each frame with its corresponding coordinates. During the object detection phase, feature vectors are obtained using 2D correlation, a statistical method that directly measures the similarity between two video frames, irrespective of changes in lighting conditions or object translations. However, this technique cannot handle image scaling and rotation. CNN popularized a two-stage technique in which a classifier is used to a sparse set of potential object locations to produce highly accurate object detectors. One-stage detectors, on the other hand, are used over a regular and dense sampling of potential item locations, making them more efficient and uncomplicated but less accurate than two-stage detectors. To train the deep learning neural networks to be correlated with specific objects, their classification, attributes, and training data are required. In the case of video analytics, the video must be divided into individual frames to identify each object or item. Video analytics can be used to analyze video evidence, such as identifying individuals wearing pink pants, using algorithms. This study proposes the use of computer intelligence (CI) and artificial intelligence (AI) algorithms to detect intelligent motion and compares their respective performances. CI and Al are the two most dominant technologies in technical society.