MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641062238 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology, Hyderabad, Telangana, on May 16, for 'an edge surveillance apparatus for threat detection from live video streams.'

Inventor(s) include Uppala Suhas; Shruthika Sunku; M. Khyathi Meghana; Tridandapani Vijaya Soumya; and Nishanth Koganti.

The application for the patent was published on May 29, under issue no. 22/2026.

According to the abstract released by the Intellectual Property India: "The invention is a live RTSP-based edge surveillance pipeline deployed on the NVIDIA Jetson Orin NX edge device with 16 GB RAM. This integrated real-time system performs both violence detection and weapon detection locally on live CCTV video streams accessed via RTSP URLs. The violence detection module employs a 3D Convolutional Neural Network (3D CNN) trained from scratch to analyze spatio-temporal patterns of human behavior, while the weapon detection module utilizes the YOLOv11 architecture for precise identification of weapons such as guns and knives.The system processes video streams entirely on the edge device, minimizing latency and bandwidth usage by sending only alert metadata and relevant evidence to a remote server. If the server becomes unavailable, the system retains alert evidence locally to ensure no loss of critical data. The alert logic incorporates confidence thresholds and requires verification of human presence in the frame before triggering violence alerts, enhancing detection reliability.The edge deployment is optimized through ONNX model conversion to ensure efficient inference on the Jetson Orin NX hardware. A major practical contribution of this invention is the avoidance of continuous raw video streaming to the cloud, thereby improving privacy and reducing network load. The system was validated on a campus-collected dataset for both weapon and violence detection, with measured alert delay verified to be less than one second, supporting real-time threat response."

Disclaimer: Curated by HT Syndication.