STREAM PROCESSING ALGORITHMS FOR DYNAMIC 3D SCENE ANALYSIS
Abstract
The goal of this effort was to take the computation closer to the sensor when it made sense to do so. Wide Area Motion Imagery (WAMI) data was used given the high bandwidth requirements associated with WAMI. Our focus was on; feature detection and matching techniques along the video sequence; developing efficient structure-from-motion and bundle adjustment techniques for high resolution aerial imagery; designing 3D reconstruction algorithm to run on metropolitan-scale data; analytical image stabilization and geo-registration; and developing moving object detection and tracking algorithms. The creation of 3D models from airborne wide area motion imagery (WAMI) will facilitate a number of capabilities including improved geo-registration, increased reliability of target tracking through occlusions and shadows, onboard processing for video object analysis, better modeling of object motion dynamics, and enable high compression of large scale scene imagery for efficient real time downlinks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 15, 2018
- Accession Number
- AD1047175
Entities
People
- Kannappan Palaniappan
Organizations
- University of Missouri