AI-Driven Human Behavior Analysis for Smart Surveillance Using Computer Vision and Explainable AI

Abstract

The fast development of intelligent surveillance systems has enhanced the need to have intelligent and dependable analysis of human behaviour through computer vision methods. To overcome these obstacles, this paper presents an AI-enabled platform for analyzing human behavior in intelligent surveillance settings with the help of computer vision and explainable artificial intelligence (XAI). In the proposed model, a hybrid deep learning system is used based on Convolutional Neural Networks (CNN) to extract spatial features and Long Short-Term Memory (LSTM) networks to model behavior over time. Moreover, explainability tools like Local Interpretable Model-Agnostic Explanations (LIME) and Shapley Additive exPlanations (SHAP) are included to give local and global interpretability of model predictions to increase transparency and trust. It is tested on benchmark human activity datasets and performs better than both traditional machine learning and baseline deep learning models, with an accuracy of 94.6. The proposed model also has better precision, recall, and F1-score, and is reliable for detecting complex human behaviors. Also, the explainable AI implementation allows more thorough observation of significant behavioral characteristics (motion pattern, posture, and interaction dynamics). The findings suggest that the suggested framework provides an effective, precise, and understandable solution to real-time smart surveillance applications.