MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074347 A) filed by Vellore Institute Of Technology on June 16, 2026, for Adaptive Low-Light Video Surveillance System With Multi-Modal Anomaly Detection.
Inventors include Dr Perepi Rajarajeswari; Mallidi Sai Jaswanth Reddy; and Harshitha Gangavarapu.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT ADAPTIVE LOW-LIGHT VIDEO SURVEILLANCE SYSTEM WITH MULTI-MODAL ANOMALY DETECTION A method (100) for integrated adaptive-luminance cognitive surveillance includes ingesting (102) a raw video stream from a surveillance camera under low-light conditions and preprocessing (104) frames to apply noise reduction. The method (100) includes adaptively enhancing (106) preprocessed frames using a deep learning model to adjust brightness and contrast while preserving structural details including edges. The method (100) includes extracting (108) features from enhanced frames along three parallel paths comprising a spatial path (214) for object detection, a motion path (216) for optical flow estimation, and a temporal path (218) for behavior sequence tracking. The method (100) includes fusing (110) spatial, motion, and temporal features using a multi-modal fusion module to generate a unified feature representation, computing (112) an anomaly score by comparing the unified representation against learned normal behavior patterns, and generating (116) a real-time alert when the anomaly score exceeds a configured threshold.
Disclaimer: Curated by HT Syndication.
