Image Matching for Stabilizing Airborne Imagery with Shallow Water Applications

Abstract

Airborne imaging techniques are commonly used for observation of geospatial variability in terrestrial and oceanographic surface properties. The rapid, relatively large-scale sampling makes airborne imaging attractive tools for coastal change and habitat monitoring. Difficulties arise when the uncertainty in instantaneous camera orientation is high (common to most airborne inertial navigation systems, even those coupled with GPS-based positioning) or there is an unresolved latency between image acquisition and navigation. One solution to this difficult problem is to utilize fixed features identifiable in the imagery to determine the relative orientation of a sequence of video frames, and subsequently ortho-normalize the imagery such that the frames can be seamlessly overlaid upon one another. Techniques to register airborne imagery are not new, but require significant manual identification of known ground points which is entirely impractical for video data where thousands of frames require accurate registration. In this work, we discuss image matching and combinatorial optimization techniques (developed for towed cameras at sea) to stabilize airborne video imagery and construct image mosaics. The techniques are robust enough to allow for video obtained over a moving sea surface to be stabilized, provided at least some portion of the video overlooks or, perhaps, only temporarily looses sight of land (for example, when passing over an inlet). The high relative accuracy between frames allows for accurate relative geospatial registration of pixel elements to be performed; a requirement of applications requiring time series of surface optical signals or applications utilizing time-average mosaics. The key element of the image matching is automatically finding features whose relative positions can be matched, allowing one image to be adjusted relative to the other. Sequential adjustment creates a continuous, seamless video from which geospatial data can be extracted. We will discuss applications related to habitat monitoring in estuarine systems and for bathymetric inversion in shallow coastal waters. This work was sponsored by the Joint Airborne Lidar Bathymetric Technical Center for Expertise (JALBTCX) and the NOAA Office of Coastal Survey.

Department

Center for Coastal and Ocean Mapping

Publication Date

2-2010

Journal Title

Ocean Sciences Meeting

Conference Date

Feb 22 - Feb 26, 2010

Publisher Place

Portland, OR, USA

Publisher

American Geophysical Union Publications

Document Type

Conference Proceeding

Link to Full Text

Share

COinS