Using Orthophoto for Building Boundary Sharpening in the Digital Surface Model

Xiaohu Lu

Xiaohu Lu Civil, Environmental and Geodetic Engineering
Rongjun Qin Civil, Environmental and Geodetic Engineering
Xu Huang Civil, Environmental and Geodetic Engineering

13P

Nowadays dense stereo matching has become one of the dominant tools in 3D reconstruction of urban regions for its low cost and high flexibility in generating 3D points. However, state-of-the-art stereo matching algorithms normally assume the surface geometry pieces-wise planar (Hirschmuller, 2008), where a smooth penalty is imposed to deal with non-texture or repeating-texture areas. This on one hand, generates much smooth surface models, while on the other hand, may partially leads to smoothing on depth discontinuities, particularly for fence-shaped regions or densely built areas with narrow streets. Although ad-hoc or adaptive tuning of the smooth penalty parameters may seem to work to a certain degree while it is often a research question of how these parameters should be appropriately selected. Given that in a top-view mapping scenario, the orthophoto corrected by the produced digital surface model (DSM) often preserve building boundaries well, we therefore in this paper consider a post-process strategy that directly leverage the consistency on the building boundaries both on the DSM and orthophoto.

Related post-processing methods generally take pixel and contour/line as primitives (pixel-based and contour line based). The pixel based methods (Huang and Zhang, 2016; Park et al., 2015) normally give a support window centered at pixels on the boundaries, and utilize the disparity of neighbor intensity-similar pixels to refine the disparity of central pixel. The contour/line based methods (Hirschmuller, 2008; Li et al., 2015) refine the entire boundary set instead of individual pixels. They usually extract building boundaries or line segments first, and then use plane function to refine the disparity of these boundaries.

In this work, we propose to use a more informative primitive, 2D polygon, to improve the effectiveness of the post-process. Our strategy is to apply a segmentation algorithm to divide the entire orthophoto into a polygon sets first, then we project each polygon onto the DSM, and apply planar criteria to evaluate if the corresponding DSM represent a plane. For those planes, a 3D plane will be fitted with which the boundaries of the plane strictly follows the polygon shape in the orthophoto. Comparative results with original DSM and refined DSM of other algorithms on various satellite stereo images shows the robustness and effectiveness of the proposed DSM sharpening method.

 

12:45 Using Orthophoto for Building Boundary Sharpening in the Digital Surface Model, Xiaohu Lu

January 29 @ 12:45
12:45 — 12:50 (5′)

Mineral DEFG

Xiaohu Lu

Add to Google Calendar