Open source satellite derived bathymetry module for NASA Ames Stereo Pipeline
Abstract
The NASA Ames Stereo Pipeline (ASP) is an open source software of automated geodesy and photogrammetry tools to process satellite, aerial camera, and historical imagery with and without accurate camera pose information using a structure-from-motion (SfM) methodology. ASP is designed to generate topographic digital surface models (DSM). We added to the ASP topographic module a bathymetric module to derive near shore bathymetry using satellite panchromatic bands (PAN), and multispectral bands. The process is semiautomatic and can generate either topographic, bathymetric or topo-bathymetric (TB) seamless 3D point clouds and DSM in the same vertical and horizontal coordinate systems. The bathymetric results depend heavily on the water surface elevation and while previous methods considered the water surface horizontal, our bathymetric module takes into consideration the earth curvature for the considered satellite imagery. A land / water mask can be automatically derived using NIR bands or can be user defined. The new ASP bathymetry module was tested using WorldView-2 panchromatic and green band stereo imagery in Florida Keys (Key West) FL from May 2015. The nearshore PAN and GRN bathymetric results around Key West, FL were validated against bathymetric lidar collected in 2017. The validation errors improve with adding camera calibration and finally alignment to prior topographic lidar data (no bathymetry data used) from 1.0778 m root mean square error (RMSE) to 0.4052 m RMSE to 0.2480 m RMSE, respectively. For PAN bands the depth penetration around Key West was around 4 m with a TB DSM resolution of 1 m. For the same area, using the green (GRN) band the bathymetric validation RMSE in absence of camera adjustment or alignment was 1.1040 m, with camera adjustment RMSE improves to 0.6846 m, and with topographic alignment RMSE is 0.5854. For the green bands the depth penetration in Key West was approximately 7 m with a TB DSM resolution of 2 m.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.G54A..06P