Members:

Shawn Recker, Mauricio Hess-Flores, Mark A. Duchaineau, Kenneth I. Joy

Overview

We present a novel, user-interactive visualization framework for computer vision research, that allows for the analysis of scene structure uncertainty and its sensitivity to parameters in different multi-view scene reconstruction stages. The input to our tool is a set of input camera parameters, feature tracks and scene structure previously computed from a set of images, such as those in Figure 1. The output is combined statistical, visual, and isosurface information, which provides user insight into the sensitivity of scene structure at the stages leading up to structure computation, such as frame decimation, feature tracking, and self-calibration. The volume rendering-based approach first creates a scalar field for a chosen structure position, at a user-specified size and resolution, from angular error measurements. A selected position in red enclosed by a green bounded region is shown in Figure 2 and magnified in Figure 3, with camera positions rendered in blue. The scalar field corresponding to the bounded region, shown in Figure 4, depicts lower positional uncertainties in red and higher ones in yellow and green. The red user-defined isosurface, which contains the ground-truth structure position, depicts regions of lowest uncertainty. The visible column-like shape indicates that lower uncertainty is seen in the directions along the lines of sight of the cameras. Changes in scalar field values with changes in reconstruction parameters are indicative of sensitivity to such parameters. In summary, our tool and its user interaction allows for such an uncertainty and sensitivity analysis in ways that have traditionally been achieved mathematically, without any visual aid. Results have been demonstrated for different types of camera configurations, and we have confirmed for example how over-decimation can be detected using the proposed technique, and how feature tracking inaccuracies have a stronger impact on scene structure than the camera's intrinsic parameters.

Related Publications

• Mauricio Hess-Flores, Shawn Recker, Kenneth I. Joy, Mark A. Duchaineau, "Visualization Methods for Computer Vision Analysis", in "Fifth International Conferences on Pervasive Patterns and Applications (PATTERNS 2013)", Valencia, Spain, 2013.

• Shawn Recker, Mauricio Hess-Flores, Mark A. Duchaineau, Kenneth I. Joy, "Visualization of Scene Structure Uncertainty in a Multi-View Reconstruction Pipeline", in "Vision, Modeling and Visualization Workshop (VMV 2012)", pp 183--190, 2012.

• Shawn Recker, Mauricio Hess-Flores, Mark A. Duchaineau, Kenneth I. Joy, "Visualization of Scene Structure Uncertainty in Multi-View Reconstruction", in "Applied Imagery Pattern Recognition (AIPR) Workshop", 2012.

Return Home