MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202621074069 A) filed by Ashokrao Mane Group Of Institutions, An Autonomous Institute, Kolhapur on June 15, 2026, for An Integrated Multimodal Computer-Vision System And Method For Real-Time Disaster Structural- Damage Assessment And Victim Detection Using Deep Learning.
Inventors include Prof. Priyanka R. Jadhav; Prof. Pradnya P. Nagarale; Dr. Shrenik S. Sarade; Prof. Anupama D. Patil; Prof. Pooja A. Patil; and Mr. Aditya Mandale.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: An integrated multimodal computer-vision system (100) and method for real-time disaster structural-damage assessment and victim detection is disclosed. A common acquisition interface (110) receives a visual feed interchangeably from a webcam, UAV/drone, CCTV, or uploaded image, and normalises each frame into a shared frame supplied simultaneously to three parallel perception branches. A victim-detection branch (130) applies instance segmentation to localise human victims with masks, boxes, and confidence scores. A damage-assessment branch (140) detects building regions and computes, for each, a quantitative structural-damage score by combining a damage-segmentation measure with a multi-scale edge-analysis measure. A fire-and-smoke branch (150) localises flame and smoke by colour-space analysis. A fusion-and-scoring module (160), operating on the identical shared frame, spatially correlates victims with damaged and burning structures and computes a single composite severity-ranked situational report, rendered (170) as a real-time overlay with statistics. Unlike prior single-task systems, the invention unifies all three perception tasks on one live feed and fuses them into one prioritised, quantitatively-scored output, deployable on GPU, edge, or cloud infrastructure for emergency- response, smart-city, drone, and infrastructure-inspection applications.
Disclaimer: Curated by HT Syndication.
