Image Sequences to Image Pairs

Space situation awareness (SSA) requires the ability to monitor, identify, and catalog a wide variety of uncooperative space objects. To help improve Air Force SSA capabilities, we propose the Image Sequences to Image Pairs (ISIP) system. ISIP's input is a sequence of images from a telescope (or micro-satellite such as XSS-10/11) showing an uncooperative space object passing overhead. ISIP processes the sequence in real-time to select sets of stereo image pairs. Each pair is combined into a single anaglyph stereo image and also used for 3D stereo reconstruction. At the core of our system is a structure-from-motion algorithm that computes an estimate of the satellite's full six-degree of freedom pose, based on frame-to-frame changes in satellite appearance. This estimate is used to determine which two input images would make a good stereo pair. Note that as desirable byproducts, our technique recovers the 3D motion of the object (position, orientation, and angular velocity) and the 3D shape of the object. In Phase I, we will develop a fully functional ISIP software system in C++ and perform an evaluation using simulated sensor data of uncooperative space objects. Our Phase II effort will focus on tuning the system to space surveillance imagery and optimizing the system to increase accuracy and throughput.