MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202643077018 A) filed by Mahatma Gandhi University on June 22, 2026, for Method And System For Rare Pattern Detection In Surveillance Video.
Inventors include Mintu Movi; Dr. Abdul Jabbar P; and Prof. Dr. Bindu V R.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: The present disclosure relates to a method and system for rare pattern detection in surveillance video. The method comprises preprocessing a video file to extract a plurality of distinct image frames, segmenting image frames having at least one object, and analyzing the image frames based on visual features. The preprocessing includes computing similarity distances between consecutive image frames using histogram comparison based on Bhattacharyya distance, normalizing and scaling the similarity distances, determining an optimal similarity threshold using Otsu's method, and filtering redundant frames to obtain unique frames. The segmentation includes performing object detection using a pre-trained Detectron2 Mask R-CNN R50-FPN model, dynamically determining an optimal confidence threshold using an Elbow method, filtering detected instances, removing small-sized object instances, and refining segmentation masks using morphological operations. The analyzing includes extracting instance-level and global contextual features, generating a unified feature representation, performing Fuzzy C-means clustering with adaptive parameter selection, and assigning dynamic rarity levels to identify rare and security-relevant patterns in surveillance environments. (Fig. 1)
Disclaimer: Curated by HT Syndication.
