MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202641024712 A) filed by Sr University, Warangal, Telangana, on March 2, for 'a modified convolutional neural network based real-time anomaly detection system for intelligent video surveillance using mask r-cnn segmentation.'

Inventor(s) include Mr. Gudupudi Ravi Kumar; Dr. V. Malathy; and Dr. K. Rajkumar.

The application for the patent was published on March 13, under issue no. 11/2026.

According to the abstract released by the Intellectual Property India: "The present invention discloses a Modified Convolutional Neural Network Based Real-Time Anomaly Detection System for Intelligent Video Surveillance Using Mask R-CNN Segmentation. The system is designed to enhance automated monitoring capabilities by integrating advanced deep learning techniques for accurate object detection, instance segmentation, and abnormal behavior recognition in dynamic environments. The framework comprises a video acquisition module configured to capture continuous surveillance streams, a modified convolutional neural network backbone optimized for high-resolution feature extraction, and a Mask R-CNN-based segmentation engine for precise object localization and boundary identification.A temporal behavior analysis module tracks object movement across sequential frames to extract motion dynamics, trajectory patterns, and interaction characteristics. An anomaly detection engine processes spatial and temporal features to generate anomaly scores using supervised or semi-supervised learning models. The system dynamically adapts detection thresholds to accommodate environmental variations such as lighting changes, crowd density, and camera motion.Edge computing compatibility enables low-latency real-time inference, while model optimization techniques including pruning and quantization improve computational efficiency. The system further incorporates explainable AI mechanisms, encrypted event logging, and automated alert generation with annotated visual evidence. By combining instance-level segmentation with intelligent behavioral modeling, the invention improves surveillance accuracy, reduces false positives, enhances security responsiveness, and supports scalable deployment in smart cities, transportation hubs, industrial facilities, and critical infrastructure environments."

Disclaimer: Curated by HT Syndication.