Cross-track satellite stereo for 3D modelling of urban areas

Abstract

Stereo imagery from very high-resolution satellite series can be used to create high-resolution digital surface models (DSMs) with sub-meter spatial resolution. Especially when in-orbit tristereo acquisitions are available as provided by the Pléiades constellation, buildings can be reconstructed well even in dense city centres. Unfortunately, such high-quality stereo data are often not available in the image archives and, thus, need to be newly acquired in most cases, leading to long delays and high cost. Especially for rapid mapping or crisis-response applications, high-resolution DSMs would be helpful for tasks such as city modelling and ultimately population estimation. In this work, we test the feasibility of producing DSMs by pairing two archive images, which were not acquired on the same day, in the same orbit or by the same satellite. We tested this approach for 14 images from Port-au-Prince, Haiti, and 12 images from Salzburg, Austria, from which we calculated a total of 136 DSMs. These were evaluated against Lidar DSMs. The resulting DSMs reach a completeness of up to 60%, and partly Median Absolute Deviation values comparable to in-orbit stereo pairs. For best results, the stereo convergence angle should be between 5 and 15°, and the difference in sun angles between the two images should be below 25-30.

Evaluation of Digital Elevation Models Derived from Multi-Date Satellite Stereo Imagery for Urban Areas

“…On one hand, the convergence angle plays a decisive role. While Krauß et al (2019) indicate an optimal angle between 5 and 15 degrees for Pléiades images, Facciolo et al (2017) report an optimal angle of about 20 degrees for Worldview 3 images. On the other hand, it is obvious that the time difference between the two images is an important factor.…”

“…Often, however, high-quality in-orbit stereo image pairs are either not available in the archives, or have to be acquired in a time-consuming process (Krauß, D'Angelo, & Wendt, 2019). To address this shortcoming and reduce the reliance on in-orbit stereo pairs, several works have attempted to create DSMs from images with different acquisition times (multi-date), or combining images from different satellites (cross-sensor) (Facciolo, De Franchis, & Meinhardt-Llopis, 2017;Krauß et al, 2019;Ozcanli et al, 2015;Qin, 2019). While these works had access to a tremendous stock of images (up to 200 images per site), humanitarian organizations do not have the budget to acquire this number of images.…”

Satellite Imaging Direction Angles Estimation Method Based on Rational Polynomial Coefficients

“…Satellite imaging direction angles, including the azimuth angle and the incidence angle, are used to describe a satellite position relative to a ground object at the moment of imaging. In 3D reconstruction, a stereo pair consisting of two images with different direction angles can be used to reconstruct a 3D model of the ground object (De Franchis et al, 2014;Hu et al, 2016;Krauß et al, 2018). However, the quality of 3D models obtained from different image pairs varies considerably.…”

Multiple View Geometry in Remote Sensing: An Empirical Study Based on Pléiades Satellite Images

“…Since the Pléiades sensors are also able to steer in across track direction they can collect images over the same scene on ground from different orbits yielding also across track stereo pairs. Such photogrammetry-like multi-overlap satellite images were, for instance, processed in [2,7]. Nonetheless, the presented tri-stereo across track data is unique in literature and the underlying potential is certainly worth being investigated.…”