MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641085944 A) filed by Bvrit Hyderabad College Of Engineering For Women; and Dr Kayal Padmanandam on July 14, 2026, for Human-Centric Video Summarization With Adaptive Storage.
Inventors include Dr. Kayal Padmanandam; and Myakala Pooja.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: The widespread adoption of surveillance systems across smart cities, educational campuses, transportation facilities, and commercial establishments has led to an unprecedented increase in the volume of video data generated every day. Since these cameras continuously capture video regardless of the occurrence of meaningful activities, large portions of the recorded footage contain repetitive or inactive scenes. This results in excessive storage consumption, increased management costs, and difficulties in efficiently searching for relevant events. Traditional surveillance solutions generally archive complete video streams without distinguishing between critical incidents and non-essential content, thereby limiting their effectiveness for intelligent monitoring. To overcome these limitations, this work presents a human-centric video summarization framework designed for smart surveillance applications by leveraging the YOLOv8 object detection model. The proposed approach prioritizes the identification of human presence and behavior because human activities represent the most significant events in surveillance environments. Video streams are divided into fixed- duration segments and analyzed individually, where YOLOv8 detects people in each segment while motion analysis evaluates the level of activity. By combining object detection with motion information, the framework identifies event-rich segments that contain meaningful human interactions. These informative segments are retained in the summarized video, whereas static or redundant portions are eliminated or highly compressed to reduce storage overhead. Furthermore, the framework generates compact metadata for every summarized segment, including temporal information, the number of detected persons, motion intensity values, and event classifications. Experimental results indicate high human detection performance with substantial reductions in video storage requirements, demonstrating next-generation intelligent surveillance systems.
Disclaimer: Curated by HT Syndication.
