You are here

Automated 3D Reconstruction and Pose Estimation of Space Objects Using Ground Based Telescope Imagery


TECHNOLOGY AREA(S): Sensors, Electronics, Battlespace 

OBJECTIVE: Using a series of ground captured satellite imagery, automatically perform image registration to previous passes and simulations. Construct a 3D reconstruction of a satellite evaluating identity, pose, and configuration in less than 15 minutes. 

DESCRIPTION: Using full passes of image data which show multiple satellite orientations, automatically create 3D wireframe images and automatically compare to existing models, evaluating identity, pose, and configuration change in less than 15 minutes (goal). 

PHASE I: Develop an algorithm that automatically finds and matches features of satellite imagery to those of previous passes and 2D images produced from simulations. Register the images. 2D images of training satellite passes will be provided along with previous training images of the same satellite and 2D images from simulations. The automated process should function entirely on a standalone PC system. 

PHASE II: Produce sparse and dense point cloud reconstructions of a satellite object. A sparse point cloud can be determined using multiple images of a satellite pass. A variety of techniques should be pursued that have the ability of performing a dense reconstruction, including shape from shading. This phase will demonstrate the ability to produce a 3D reconstruction with accuracies within 500 nrad of a satellite using a series of images. 

PHASE III: Using the 3D point cloud of an object, perform a 3D registration to that of known model. This process should determine within a series of possible 3D poses, which pose is most appropriately matched to the 3D model that was derived from 2D imaging, within 15 minutes using a standalone PC. 


1. Charles L. Matson, Kathy Borelli, Stuart Jefferies, Charles C. Beckner, Jr., E. Keith Hege, and Michael Lloyd-Hart, "Fast and optimal multiframe blind deconvolution algorithm for high-resolution ground-based imaging of space objects," Appl. Opt. 48, A75-A92 (2009).

2. Michael Werth, Brandoch Calef, Daniel Thompson, Kathy Borelli, and Lisa Thompson, Recent improvements in advanced automated post-processing at the AMOS observatories, Proceedings of IEEE Aerospace, March 2015.

3. David R. Gerwe and Paul Menicucci, A real time superresolution image enhancement processor, Proceedings of AMOS, Sept. 2009.

4. Daniel Thompson, Michael Werth, Brandoch Calef, David Witte, and Stacie Williams, Simultaneous processing of visible and long-wave infrared satellite imagery, Proceedings of IEEE Aerospace, March 2015.


KEYWORDS: Computer Vision, 3D Reconstruction, Digital Image Processing, Satellite Imagery, Machine Learning 

US Flag An Official Website of the United States Government