MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070925 A) filed by Noida International University on June 08, 2026, for An Ai-Enabled Uav (drone) Based System And Method For Real-Time Highway Damage Detection Using Image Enhancement And Yolo Object Detection..
Inventors include Vimal Bibhu; and Anup Saxena.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: ABSTRACT UAV (Drone)-based highway inspection has become a focal point in the fields of intelligent infrastructure monitoring because it offers faster coverage, enhanced safety, and improved operation compared to traditional manual surveys. Nevertheless, aerial images/video captured by drones typically show non-uniform illumination, shadows, sensor noise, and low contrast. These anomalies hinder the detection of small surface defects and can lead to a poor performance of automated defect detection models. This present invention presents a drone-based system for highway defect detection that overcomes these limitations. Before the training of the detection model, the quality of the aerial images is ameliorated by two preprocessing steps including CLAHE and bilateral filtering for a local contrast enhancement and noise reduction while edge-information can be kept. After that the edge enhancement and clarity feature will improve the feature clarity of defects. Thus, the improved images were used for training a YOLO based object detection model. A custom dataset consisting of 3,344 annotated highway images was employed for model training and validation. We evaluated the model using widely accepted detection metrics namely precision, recall, F1-score and mean Average Precision (mAP) at various IoU thresholds (mAP@50 and mAP@50-95). The trained model presents a precision value of 0.936, a recall of 0.948, an F1-score of 0.94, mAP@50 is 0.975 and mAP@50-95 is 0.7489. This implies the convergence of training is stable, the ability to classify many categories of defects is strong, it is clearly observed that contrast enhancement and edge-preserving filtering significantly improve the reliability of UAV-based highway defect detection.
Disclaimer: Curated by HT Syndication.
