MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641090172 A) filed by Kalaimathi Bathirappan; Vijayakumar Kandasamy; Anusulatha Karunanithi; Adline Jancy Yesu Arumaidhas; Rajeshwaran Kandhasamy; Chandru Ramasamy; Nithya Nagamanikkam; and Jayanthisree Sundaram on July 24, 2026, for Multi-Scale Deep Learning Approach For Robust License Plate Detection Under Challenging Environmental Conditions.
Inventors include Kalaimathi Bathirappan; Vijayakumar Kandasamy; Anusulatha Karunanithi; Adline Jancy Yesu Arumaidhas; Rajeshwaran Kandhasamy; Chandru Ramasamy; Nithya Nagamanikkam; and Jayanthisree Sundaram.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: The present invention discloses a multi-scale deep learning system for robust license plate detection under challenging conditions including motion blur, skewed camera angles, and physically degraded plate surfaces. The system comprises a camera input subsystem acquiring vehicle image frames, operatively coupled to an image preprocessing module applying adaptive deblurring, contrast enhancement, and geometric normalization. A multi-scale convolutional neural network backbone generates hierarchical feature maps at multiple spatial resolutions, enabling plate detection across varying distances and sizes. A detection and localization module generates bounding box coordinates, confidence scores, and orientation estimates, while a post-processing module applies non-maximum suppression and geometric correction to produce refined plate region crops. An output interface transmits detection results to connected downstream systems. The system provides improved detection reliability, enhanced robustness under physically degraded conditions, reduced false detection rates, optimized real-time processing efficiency, and scalable deployment across intelligent transportation, traffic enforcement, and vehicle access control applications.
Disclaimer: Curated by HT Syndication.
