Large-scale surveillance networks produce overwhelming volumes of video data that are difficult to analyze in real time, especially when operators have only general descriptions rather than precise identifiers. Centralized systems that transmit raw video create latency, scalability limits, and privacy concerns, while supervised biometric tools such as facial recognition are insufficient for vague queries and raise ethical issues. As a result, current systems struggle to provide efficient, privacy-preserving search capabilities for mission-critical environments.
The invention is an interactive surveillance system operating on an edge–fog–cloud architecture. Edge nodes perform human pose estimation to locate keypoints and crop body regions before extracting initial color features. Fog nodes receive only these features and apply clustering to determine dominant colors and map RGB centroids to human-readable labels through a predefined dictionary. Operators issue high-level queries such as clothing color, which are matched against extracted features to return relevant frames and camera IDs in under two seconds. The system uses a containerized microservices design for scalability and minimizes privacy risk by avoiding transmission of raw video.
• Enables real time querying using high level semantic descriptions
• Preserves privacy by eliminating raw video transmission
• Reduces network load through edge based feature extraction
• Improves scalability with microservices architecture
• Delivers sub two second query response for mission critical use
• Operates effectively on low cost edge hardware
• Supports decentralized processing for system resilience
• Allows flexible searching without biometric identifiers
• United States 12,211,277 - Issued 01/28/2025
Prototype
This technology is available for licensing.
Strong potential for adoption by security system providers, smart city developers, and infrastructure operators seeking real-time, privacy-preserving video analytics with reduced bandwidth and scalable edge deployment for mission-critical environments.
Prototype system demonstrations and performance validation details available upon request.