Dynamic Pushbroom Stereo Mosaics for Moving Target Extraction

Our goal is to rapidly acquire panoramic mosaicing maps with information of all 3D (moving) targets as a light aerial vehicle with a video camera fights over an unknown area. Parallel-perspective pushbroom stereo mosaics approach has been proven to be an effective method to stabilize image sequences and to obtain better 3D reconstruction results than using interframe motion stereo. In this paper we propose a dynamic pushbroom stereo mosaic approach for representing and extracting independent moving targets from unknown 3D scenes. Each mosaiced image (obtained from time-varying images) is virtually a “moving picture”, and multiple such mosaics provide salient motion features from various viewing directions. Independent moving targets can be easily identified in the matching process of stereo mosaics by detecting “out-of-place” regions, which violate epipolar constraints and/or give 3D anomaly. We propose a segmentation-based stereo matching approach with natural matching primitives to estimate the 3D structure of the scene, particularly the ground structures (e.g., roads) on which humans or vehicles move, and then to identify moving targets and to measure their 3D structures and movements.