MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106686 A) filed by Raghava M S; Dayananda Sagar Academy Of Technology And Management; Aviral Sahu; Dheeraj A U; Kushal N Seety; and Shashwat Yadav on September 04, 2026, for Real-Time Ai-Based System And Method For Detecting Loitering And Unattended Objects In Public Surveillance Networks Using Detector– Tracker Fusion And Dwell-Time Analysis.
Inventors include Raghava M S; Dayananda Sagar Academy Of Technology And Management; Aviral Sahu; Dheeraj A U; Kushal N Seety; and Shashwat Yadav.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present invention discloses a real-time, edge-deployable surveil-lance system for automatically detecting loitering individuals and unattended (abandoned) objects in public spaces such as airports and metro stations. The system ingests live CCTV video and, prior to any analysis, applies a privacy pre-filter that blurs the faces of individuals who have not been flagged by the downstream logic. A lightweight object detector (e.g., a YOLO-family network) identifies persons and objects in each frame, and an identity-persistent tracker (e.g., a Kalman-filter and assignment-based tracker) maintains a con-sistent identifier for each person or object across frames despite brief occlusions. A restricted-zone and ownership-geometry module de-fines monitored areas as configurable polygons and assigns object ownership to the nearest person within a defined radius. A dwell-time analysis module, smoothed by a lightweight recurrent (LSTM) layer to bridge short tracking gaps caused by occlusion, computes the elapsed time a tracked identity remains within a zone (for loi-tering) or the elapsed time an owner remains separated from their object (for abandonment). When either dwell-time threshold is ex-ceeded, the system automatically generates and transmits an alert, via an application programming interface, to a security desk, includ-ing the camera identifier, zone identifier, elapsed dwell time, and a short video clip of the event. The invention reduces reliance on con-tinuous human monitoring, lowers false-alarm rates common in dense crowds through proximity-and-displacement ownership logic, and preserves the privacy of unflagged individuals through selective face-blurring and limited full-resolution retention.
Disclaimer: Curated by HT Syndication.
