Skip to end of metadata
Go to start of metadata

New Toolbox will aid in remotely sensing river bathymetry

USGS Scientist and SurfBoard member, Carl Legleiter has release a new toolset for processing multi-spectral imagery of shallow rivers. The Optical River Bathymetry Toolkit, ORByT, is an publicly available application to facilitate mapping water depth in river channels from passive optical image data. ORByT provides an integrated workflow for image processing, depth retrieval, and export of results.

Example river bathymetry output from the ORByT

Carl has done an excellent job of building a nice code base over on the USGS Gitlab, and you can download the Toolkit, documentation, and training here:

https://code.usgs.gov/cjl/orbyt-optical-river-bathymetry-toolkit

There is even a DOI Talent self-lead course (this is available to DOI employees only) on the usage of the Toolkit: 

https://doitalent.ibc.doi.gov/course/view.php?id=15380


Citation

Citation:

Legleiter, C.J., 2020, ORByT - Optical River Bathymetry Toolkit (ver. 1.1, June, 2020): U.S. Geological software release, https://doi.org/10.5066/P9GXRJ3K.

  • No labels

6 Comments

  1. very cool Carl! I will make sure to take a closer look and share with a few colleagues in Canada.

    What exactly is "passive optical image data" and what are the specific image input requirements for this tool to produce reliable water depth?

    Liz



    1. Thanks for your interest, Liz, please do let your colleagues know about ORByT.  The term passive optical image data refers to remotely sensed data that record reflected solar radiation, as opposed to an active sensor that emits its own energy signal, such as lidar.  A standard photo from a conventional RGB camera would be an example of input suitable for ORByT.  Multispectral images from satellites, airborne hyperspectal images, or footage from a drone equipped with a video camera would all be potential inputs to the software.  The more important consideration influencing the reliability of depths estimated with ORByT is the river itself, as the method is best-suited for application to relatively shallow, clear-flowing streams.  This type of passive optical depth retrieval is not likely to yield accurate results in larger, deeper, and more turbid rivers.  Water clarity is the main constraint, so a bigger river could work, or at least be worth trying.  The eye test is relevant here - if you can see the bottom yourself, it's probably worth giving ORByT a shot.  

      Hope this helps to clarify, thanks again, and take care,


      Carl

      1. Engel, Frank AUTHOR

        Apologies for implying that the ORByT was only for multispectral imagery in my description. I'll be working myself through the excellent DOI Talent course you produced Legleiter, Carl J. to learn more. I was wondering, what level of image quality do you think you would need at a minimum to get a valid result? We obviously have several webcams aimed at rivers, that often during low flow are non-turbid and the bottom is visible. We can certainly test out both the ability of the ORByT to discern depths from RGB webcams, as well as the impacts a non-nadir image may have with some of the data we are already collecting as a part of the Image Velocimetry work we do.

        1. Thanks, Frank, glad to hear you're taking the ORByT course through DOI Talent; I appreciate the positive feedback.

          As for image quality, you don't necessarily need anything special to obtain some useful depth information.  I've shown (https://doi.org/10.1002/rra.2560) that publicly available data from the USDA's National Agricultural Imagery Program (NAIP) can yield decent bathymetric maps and my colleague Overstreet, Brandon Tylerwith the Oregon Water Science Center has had encouraging results for statewide orthos.

          I haven't tried data from webcams, but I think it's worth trying, as I am currently working with data from a Zenmuse camera commonly deployed from drones and I suspect the webcam is a similar level of quality.  You would need to rectify the image prior to input to ORByT and have some direct field measurements of depth for calibration and validation, of course, but this approach might work well.  Please keep me posted and I'd be happy to assist if I can be of help.

          Take care,

          Carl

  2. Very helpful, thanks Carl. And what about camera angle? Does your method only work for nadir images that do not require any rectification? I suppose rectification and image scaling (if necessary) could be carried out prior to use of ORByT, so this isn't necessarily a limitation?

    1. Sure, Liz, no problem.  Camera angle isn't necessarily a limitation, but I have always thought of applying the ORByT workflow to nadir images.  If you have oblique images, you would need to rectify and scale them prior to input to ORByT.  Please let me know if you try this approach, as I have never attempted to infer depth from non-nadir images.  It's worth a shot, but the different illumination and viewing geometry might cause some problems due to refraction effects and less light entering and leaving the water column at angles farther from nadir.  I'm curious ...