Intelligent Anti-Money Laundering System: An Improved YOLO-Based Suspicious Behavior Detection Framework for Banking Surveillance

With the increasing sophistication of financial crimes and money laundering activities, traditional manual monitoring methods can no longer meet the demands of modern banking security. This study aims to develop an intelligent security monitoring system for financial institutions using artificial intelligence technology. We propose an improved YOLO model-based action recognition system specifically designed for bank ATM and branch scenarios, which incorporates separable convolutions, a bidirectional feature fusion mechanism, and a spatiotemporal dual-stream network for action recognition. A rule engine is integrated to analyze suspicious behaviors. In practical applications across 20 bank branches with 436 cameras, the system achieved 87.9 % accuracy in suspicious behavior detection with an average 2.3 -second early warning time, while reducing false negatives by 56% and labor costs by 60% compared to traditional monitoring methods. These results demonstrate that the proposed system provides an effective technological solution for intelligent security monitoring in financial institutions, with significant implications for enhancing anti-money laundering efforts and banking security.