MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641061025 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on May 14, for 'system and method for early prediction of traffic accidents from video streams.'
Inventor(s) include Islabudeen M; Shreya Gupta; and Ananya Kaushal.
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: "A system (100) for early prediction of traffic accidents from video streams is disclosed. The system (100) comprises a processor (102), a memory (104) storing executable instructions, and a video input interface configured to receive a continuous CCTV video stream. The processor (102) is configured to extract and preprocess frames by resizing and normalizing pixel values, detect and track a plurality of traffic agents across frames to generate temporal trajectories, and derive motion features including spatial position and velocity. The motion features are encoded to generate latent representations. Based on the latent representations and interaction features, the processor (102) generates a plurality of future trajectories using a diffusion-based sampling process. Interactions between traffic agents are evaluated, and collision risk values are computed for each trajectory based on relative motion, proximity, and time-to-collision. A subset of highest risk values is selected using a top-k selection and aggregated to generate a final risk score."
Disclaimer: Curated by HT Syndication.
