Integrating Machine Learning and Swarm Intelligence for Enhanced Incongruity Recognition in Dense Crowds

Autonomous behavior detection in surveillance footage and the proliferation of monitoring networks are two areas where smart city and public safety initiatives are placing a premium on technological advancements. Using context modeling, new methods for object movement recognition have been developed. Sound and background movement are let in because current anomaly detection algorithms only consider geographical data and completely or partially ignore temporal data. The huge number of people in a scene causes noticeable variations in movement and appearance, making it difficult to monitor individual subjects.An technique to anomaly detection is created using machine learning along with swarm intelligence. This approach extracts spatiotemporal information from video sequences. Effective tracking and detection of anomalous items is achieved by extracting prominent characteristics from the frames in this technique. The first step is to build a two-dimensional variance plane that can capture the local spatiotemporal fluctuations within each of the pixels in a video frame. The most important areas are located using motion data in the two-dimensional variance plane and the Enhanced Particle Swarm Optimization method. The selected important video pixels are then subjected to a GLCM. In order to acquire highlevel info for anomalous event identification with more precision, a classifier based on an Enhanced Neural Network is used. Tracking accuracy suffers in low-quality footage and when objects are obstructing the view, however the anomaly detecting process is very clear.