Automated Extraction of Geospatial Features from Satellite Imagery: Computer Vision Oriented Plane Surveying

The paper explores and assesses the potential uses of high resolution satellite imagery and digital image processing algorithms for the auto detection and extraction of geospatial features, farmlands, for the purpose of statutory plane surveying tasks. The satellite imagery was georectified to provide the planar surface necessary for morphometric assessments followed by integrated image processing algorithms. Precisely, Canny edge algorithm followed by morphological closing as well as Hough transform for extracting lines of features was used. The algorithms were tested using Quick bird satellite imagery and in all cases we obtained encouraging results. This shows that computer vision and image processing using high resolution satellite imagery could be used for cadastration purposes, where property boundaries are needed and used for compensation purposes and other statutory surveying functions. The error matrix of the delineated boundaries is estimated as equal to 73.33%.