Integrating Proactive Artificial Intelligence with CCTV Cameras for Pre-Crime Criminal Intent Detection Through Body Language Analysis

Traditional CCTV surveillance operates reactively, recording events for after-the-fact review and relying on error-prone manual monitoring. This limitation raises a fundamental question: can behavioural cues be analysed automatically to identify pre-crime situations? This paper presents a systematic review of AI-driven proactive CCTV systems, examining three core layers: perception (body language, skeletal motion, eye micro-movements), analysis (transformers, GCNs, hypergraphs), and decision (edge computing, federated learning). Key findings indicate that large-scale deployments such as China's nationwide camera network can reduce crime by up to ۱۰% with a cost of approximately $۵,۹۲۲ per prevented case. However, predictive policing effectiveness remains contested, with only ۳.۷% of studies involving real-world field applications. Algorithmic bias is severe: facial recognition misidentifies darker-skinned individuals ۳۴% more often, and biased arrest data create self-reinforcing over-policing of minorities. Legal frameworks like the EU AI Act contain broad law-enforcement exemptions and exclude immaterial harms from compensation. Privacy-preserving techniques—encrypted video, skeleton extraction, federated learning, differential privacy—offer promising countermeasures but lack legal mandates. The review concludes that proactive CCTV systems must be governed by necessity, proportionality, transparency, contestability, and accountability, with independent audits, bans on indiscriminate real-time facial recognition, and community-centred oversight to ensure that technological innovation serves public safety without eroding civil liberties.

کلیدواژه ها:

Proactive CCTV ، video anomaly detection ، body language analysis ، predictive policing ، algorithmic bias ، privacy preservation