Eric D. Lester
Electrical Engineering
M.S. 1998

Feature Extraction, Image Segmentation, and Surface Fitting: The Development of a 3D-Scene Reconstruction System

Research objectives:
This research addresses the issue of building photo-realistic models of a 3D-scene from laser range sensor data. This research extracts relevant features, such as edges and corners, from both range and amplitude images to segment data into regions. The segmentation results are to be processed by fitting models to the data. Fitted models are to be used to build a photo-realistic model of the scene to be displayed for the purpose of scene characterization.

Methodology and Results:
A scene reconstruction system that takes a registered pair of range and amplitude images


and produces 3D models of objects in a scene has been built. This system incorporates a series of steps. It begins by extracting features from both range and amplitude images to create feature maps and then fusing these feature maps into a single edge map. It ends by segmenting that edge map, fitting surfaces to the objects in the scenes, and finally generating a 3D model of the scene. The feature extraction method locates both step edges and crease edges from both amplitude and range images. Fusion in the form of Bernoulli's Rule of Combination is performed on the resulting feature maps. Image segmentation is performed using a morphological watershed algorithm. Finally 3D planes, spheres, and cylinders are fit to regions in the segmented scene by the minimization of least-squared error process. The result is a system that can start with data from a laser range finder and produce, with virtually no user interaction, 3D models of objects in the scene. Results are shown for the reconstruction of scene from synthetic data, as well as real range data of actual scenes in our lab and at Oak Ridge National Laboratory.

This work was conducted by Eric D. Lester while at IRIS lab under the supervision of R. Whitaker (Thesis Chair) and M. A. Abidi. This work was supported by DOE's University Research Program in Robotics under grant DOE-DE-FG02-86NE37968.

  • Papers

    E. D. Lester, R. T. Whitaker and M. A. Abidi "Feature Extraction, Image Segmentation, and Scene Reconstruction", SPIE Conf. on Sensor Fusion and Decentralized Control in Autonomous Robotic Systems, pp. 250-260, Pittsburgh, PA, October 1997.