Application of remote sensing and UAV data for detecting unregistered and incorrectly registered real estate objects

Abstract

The completeness of the Unified State Register of Real Estate remains uneven: a share of buildings and land plots existing on the ground has no registration record, while for another share the data on contour, area or designated use diverge from reality beyond the statutory tolerance. The study quantitatively compares three sources of spatial data (submetre satellite imagery, aerial photography from an unmanned aerial vehicle, and airborne laser scanning) against two separate criteria: completeness of object detection and conformity of contour planimetric accuracy to the cadastral tolerance. An author's computational detection model was built over a 12.4 km² test site with a reference set of 1,463 capital construction objects. UAV aerial photography with ground geodetic control yielded an error of 0.072 m with 93,5% of points within tolerance; submetre satellite imagery proved unsuitable for fixing the contour (0.41 m); laser scanning provided the highest detection recall (over 0.97) at a planimetric accuracy beyond tolerance (0.11 m). In total, 28,8% of the site's objects fall outside correct registration. The optimum is reached by separating the functions of detection and coordinate fixation between sources with differing properties of the survey material rather than by selecting a single universal source.