Detecting Crime Anomalies in Smart Cities with Sharkprey Optimization and Ensemble Machine Learning

Abstract

The advent of smart cities has enabled the creation of automated criminal anomaly detection systems, owing to the copious urban data streams. This study explores the use of machine learning techniques to evaluate and identify abnormal patterns in criminal incidents. This technology improves proactive law enforcement methods, streamlines resource allocation, and helps create safer and more secure urban environments by using real-time data from several sources.In the first stage, video data is gathered from a network of strategically placed surveillance cameras across the intelligent urban area. The automated criminal anomaly detection system may be trained and improved by using this comprehensive video collection, enabling it to accurately identify and distinguish between normal and abnormal behavior patterns in varied urban settings. The collected data is subjected to preprocessing utilizing Video-to-Frame Conversion, Non-Local Means (NLM), and contrast stretching approach. The Sobel edge detection technique is used to discover the Regions of Interest (ROI) inside the frames for the purpose of segmentation, using the pre-processed data. This method integrates the White Shark Optimizer with Osprey optimization technique. Create an innovative ensemble machine learning approach to identify criminal abnormalities by combining the K-Nearest Neighbors, Random Forest, and optimum Artificial Neural Network models. To improve the precision of detection, the weight of the Artificial Neural Network (ANN) is fine-tuned using the Sharkprey Optimization approach. MATLAB is used for the execution.