GeoDrishti: Automatic Road Extraction and Alert Generation for New Roads from Satellite Imagery

Efficient monitoring of road networks is crucial for urban planning and infrastructure development. This paper presents GeoDrishti, an innovative system for automatic road extraction and alert generation from satellite imagery. Leveraging medium-resolution images from the Resourcesat LISS IV sensor available through the Bhoonidhi portal, our system addresses the challenge of detecting roads with widths of 20m and above. We employ a U-Net architecture optimized for road segmentation, coupled with a novel change detection algorithm to identify newly constructed or modified roads. The system integrates these components with an automated alert mechanism, notifying users of significant changes via email. Our implementation includes a user-friendly GUI for specifying areas of interest and managing alerts, with outputs in the form of geographically referenced shapefiles. Experimental results demonstrate an F1 score of 0.8 in road detection across diverse geographical areas, with a 90% precision in identifying newly constructed roads. The system processes a 240x240 pixel image in under 35 seconds on standard hardware, representing a 75% reduction in time compared to manual interpretation. GeoDrishti offers a scalable, efficient solution for road network monitoring, with potential applications in urban planning, map updating, and rapid response scenarios. This research contributes to the field of remote sensing by combining deep learning-based road extraction with efficient change detection, addressing the dual challenges of accurate road detection and timely change identification.