Geospatial Data Extraction from Satellite Imagery for Road Construction Planning and Monitoring

Abstract

Conventional road survey methods such as Total Station and Differential GPS (DGPS) are accurate but time-consuming, labour-intensive and spatially-limited. Thus, they are unsuitable for large-scale or rapid monitoring of infrastructure. This study speaks for the need of a scalable, affordable alternative. Using state-of-the-art satellite imagery, computer software was developed in Matlab and GIS for achieving a semi-automated structured methodology. The workflow involved taking images, geo-referencing them, converting to grayscale, noise reduction through filtering with a median filter, Canny edge detection for road boundary detection, a morphological operation by skeletonizing to extract the centreline, and finally calculating the width and length of the road programmatically. Road type classification was done using Indian Road Congress (IRC) width standards. Findings: The methodology was validated against DGPS ground-truth data for straight and curved road stretches. Accuracies of the extracted parameters were 90% for both segment types. For straight road, length was 270.12 m (accuracy: 90.00%), width was 10.90 m (90.08%). For curved road, arc length was 90.00 m (90.91%), width was 12.00 m (90.91%). To conclude, the satellite imagery-based approach is feasible, reliable, repeatable, and cost-effective tool that can supplement or replace traditional surveys. Further, this method may be implemented directly in the construction planning phase for material estimation, computation of earthworks volume, project planning, and scheduling.