Spray Paint Image Synthesis in Surveillance Camera Anomaly Detection

Smart surveillance videos system taken in unconstrained environments can be tampered with due to different environmental factors and malicious human activities. They often blur the video content and introduce difficulty in identifying the events in the scene that leads to wrong real-time decision.

• The system has a modified residual network incorporated with color attention module that significantly outperforms the state-of-the-art networks while keeping the model size small for edge applications
• A diverse background scenes to cover a wide range of application conditions of surveillance camera systems image dataset, namely “Blurveillance”, is purposedly built to be practical for application. It collects 10,000 surveillance images of five classes: "Normal", "Foggy", "Defoucs Blur", "Dirt Blur" and "Spray Paint". They cover the diverse blur anomalies caused by intentional camera tampering, unintentional factors and environmental changes
• A spray paint model, which is based on the spray paint characteristics and the thin lens model is created for synthesizing the spray paint effect

• Enables AI real-time diagnostic health/anomaly detection management of the survelliance video anomaly system
• Automatic real time signal for anomaly detection reduces man-hour for monitoring/inspecting the system status to nil
• Reduces field service, video surveillance down time and the cost of repairing and maintenance

• Surveillance camera system with large number of cameras
• Smart lamppost
• Automatic guided vechicle

• Hong Kong Short Term Patent No. HK30076502