High Resolution Surface Reconstruction from Multi-view Aerial Imagery

Abstract

Abstract-This paper presents a novel framework for surface reconstruction from multi-view aerial imagery of large scale urban scenes, which combines probabilistic volumetric modeling with smooth signed distance surface estimation, to produce very detailed and accurate surfaces. Using a continuous probabilistic volumetric model which allows for explicit representation of ambiguities caused by moving objects, reflective surfaces, areas of constant appearance, and self-occlusions, the algorithm learns the geometry and appearance of a scene from a calibrated image sequence. An online implementation of Bayesian learning precess in GPUs significantly reduces the time required to process a large number of images. The probabilistic volumetric model of occupancy is subsequently used to estimate a smooth approximation of the signed distance function to the surface. This step, which reduces to the solution of a sparse linear system, is very efficient and scalable to large data sets. The proposed algorithm is shown to produce high quality surfaces in challenging aerial scenes where previous methods make large errors in surface localization. The general applicability of the algorithm beyond aerial imagery is confirmed against the Middlebury benchmark.

Efficient dense reconstruction using geometry and image consistency constraints

“…Figure 4 shows a reference image from the Brown Site 22 dataset [44], and a dense reconstruction of the scene produced by our method. As shown in the figure, our result is noisy and unable to accurately capture many details in the scene.…”

Geo-localization using Volumetric Representations of Overhead Imagery

“…3a. This open-source GPU implementation has recently enabled practical exploitation of PVR for applications such as classification of scene objects , mesh fitting (Calakli et al 2012), video compression and tracking (Ulusoy et al 2013). In this paper, the same GPU implementation has been utilized as part of a practical geo-localization application that operates with ROIs of sizes in thousands of km 2 .…”

3D Multi-view Stereo Modelling of an Open Mine Pit Using a Lightweight UAV

“…Recent significant strides in 3D model reconstruction using Structure from Motion Multi-View Stereo-photogrammetry (SfM-MVS) algorithms in computer vision (Gong and Wang, 2011;Calakli et al 2012) is rapidly spreading on morphology based research in various scales. Production of 3D point cloud data are no longer limited to expensive and specialised devices like laser/radar range scanners, but also possible with uncalibrated consumer market digital cameras (Furukawa and Ponce, 2007;Gong and Wang, 2011;James and Robson, 2012;Bemis et al 2014).…”

“…Several interesting and useful examples from different research areas (some among many examples are Gimenez et al 2009;Niethammer et al 2010;James and Varley, 2012;Skarlatos and Kiparissi, 2012;Calakli et al 2012;Tuffen et al 2013;Bemis et al 2014;Forte, 2014;Bennet, 2015;Burns et al 2015;Haukaas, 2015;Shahbazi et al 2015;Van Damme, 2015;Vepakomma et al 2015;Tonkin et al 2016) are shaping, contributing and defining the limitations of the new modelling technique. James and Robson (2012) demonstrated the application of SfM-MVS modelling from decimeter to kilometer scale and presented a thorough discussion on the precision of the method.…”