Pixel-based approach for building heights determination by SAR radargrammetry

Abstract

ABSTRACT:Numerous advances have been made recently in photogrammetry, laser scanning, and remote sensing for the creation of 3D city models. More and more cities are interested in getting 3D city models, be it for urban planning purposes or for supporting public utility companies. In areas often affected by natural disaster, rapid updating of the 3D information may also be useful for helping rescue forces. The high resolutions that can be achieved by the new spaceborne SAR sensor generation enables the analysis of city areas at building level and make those sensors attractive for the extraction of 3D information. Moreover, they present the advantage of weather and sunlight independency, which make them more practicable than optical data, in particular for tasks where rapid response is required. Furthermore, their short revisit time and the possibility of multi-sensor constellation enable providing several acquisitions within a few hours. This opens up the floor for new applications, especially radargrammetric applications, which consider acquisitions taken under different incidence angles. In this paper, we present a new approach for determining building heights, relying only on the radargrammetric analysis of building layover. By taking into account same-side acquisitions, we present the workflow of building height determination. Focus is set on some geometric considerations, pixel-based approach for disparity map calculation, and analysis of the building layover signature for different configurations in order to determine building height. * Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.

Extraction of building parameters by SAR radargrammetric analysis of layover areas

“…Also, we apply a matching criterion that takes into account the image statistics, and thus improves the matching localization of geometric patterns (Subsection 2.3). This algorithm shows improved results compared to our previously presented approach [4] and to standard radargrammetric approaches (Section 3).…”

“…As the result is still noisy, we perform an adaptive filtering on the disparity map that allows to smooth the disparity values along the building orientation. In [4], we showed that a building is characterized in the disparity map by two parallelograms, one around the building corner, and another around the layover edge. Using a method based on standard deviation and median search (Subsection 2.4), these two parallelograms can be extracted.…”

“…Figure 1 shows a schematic representation of our workflow for building parameter extraction by SAR radargrammetry. This approach belongs to a more general approach for detecting changes by fusing InSAR and stereo-radargrammetric data [4], so that the information marked in dark blue are as- sumed to be known for the radargrammetric processing by previous computation. In our methodology for retrieving 3D information, the slave image is first coregistered and resampled on the master image (done manually until now) in slant range geometry.…”

“…As said previously and explained thoroughly in [4], our approach yields to a building signature characterized by two parallelograms of the same size in the disparity map, keeping corner and layover border distinct. Thus, we can solely derive local building height from the disparity values contained in the parallelogram that is situated around the layover border.…”

“…Due to strong geometric distortions and neighboring effects in urban areas, matches and derived building parameters are often erroneous. In a previous paper [4], we specially analyzed the results of area-based matching at building level, in order to comprehend the building signature at building level and derive correct building heights. We focused on layover areas, as they are the most reliable features for building detection in dense urban area.…”