A surface-based Next-Best-View approach for automated 3D model completion of unknown objects

Abstract

The procedure of manually generating a 3D model of an object is very time consuming for a human operator. Nextbest-view (NBV) planning is an important aspect for automation of this procedure in a robotic environment. We propose a surface-based NBV approach, which creates a triangle surface from a real-time data stream and determines viewpoints similar to human intuition. Thereby, the boundaries in the surface are detected and a quadratic patch for each boundary is estimated. Then several viewpoint candidates are calculated, which look perpendicular to the surface and overlap with previous sensor data. A NBV is selected with the goal to fill areas which are occluded. This approach focuses on the completion of a 3D model of an unknown object. Thereby, the search space for the viewpoints is not restricted to a cylinder or sphere. Our NBV determination proves to be very fast, and is evaluated in an experiment on test objects, applying an industrial robot and a laser range scanner.

Information-Gain View Planning for Free-Form Object Reconstruction with a 3D ToF Camera

“…6a and 6b. Later, Kriegel [10] used a laser scanner for modelling both the Camel and Mozart objects, providing a very precise model at the expense of a time consuming process. Fig.…”

“…Again, for viewpoint generation, the internal representation of the environment model plays an important role. Surface-based methods provide a set of viewpoints based on the location of jump edges [13], the trend of a contour [10] or the fitting of a parametric surface representation [1]. Volumetric methods provide viewpoints using the information of visited and non-visited portions of the workspace, and generally encode this space using voxel representations (or, more efficiently, octrees).…”

Appearance-based Active, Monocular, Dense Reconstruction for Micro Aerial Vehicles

“…State-of-the-art approaches to active mapping [15,5,8,25,27,22] retain only geometric information while discarding the scene appearance. As a result, a robot trying to perceive the depth of a white wall, would generate different camera trajectories in vain, eventually failing to reduce the uncertainty in the depth measurement [23].…”

“…Often, the sensor motion is restricted to a sphere and it is assumed that the object of interest is at all times located completely in the sensor frustum. Proposed algorithms reason about voxel occupancy, occlusion edges, and surface coverage [13,15]. Schmid et al [20] addressed view planning with an MAV.…”

“…The computation of the depth-map entropy, which is evaluated multiple times in every control iteration according to (15), requires summation over all pixels in the image. To maintain real-time performance, we were required to select a subset of pixels for which the information gain is computed.…”