An Automated Method for Delineating Braided River Water Surface Area from RGB Imagery
Abstract
Hydrologic research is increasingly conducted from remote sensing platforms, allowing for efficient, non-contact sampling of hydraulic parameters. In this study, a new method is presented for fully automated delineation of braided river water surfaces from Extreme Ice Survey (EIS) oblique-view RGB camera imagery, enabling rapid calculation of hydraulic parameters such as effective width (inundation area divided by reach length Smith et al., 1995; 1996) and braiding index. The test dataset is 200 high-resolution camera images obtained for the proglacial Isortoq River, southwestern Greenland, in July 2011. First, images are segmented via automated histogram thresholding, a process that also determines which images to reject for poor quality. These quality-screened images then initialize an active contouring (AC) image delineation process using the method of Li et al. (2010). This method seeks to minimize changes in contrast gradients as an initial contour determined by the automated thresholding evolves toward a final, stable solution. Comparison of this coupled thresholding/AC method with traditional supervised classification suggests that supervised classification is superior for low quality images and marginally superior for high quality images. However, this method is time consuming and impractical for long term hydrologic studies with large data volumes. Coupled AC and traditional unsupervised classifications present automated solutions for extracting hydraulic parameters, and AC proved more effective at water surface classification for the river images in this study. Results indicate that an automated solution for image selection and hydraulic parameter calculation for large data volumes can be both accurate and practical. Such a methodology will be useful in numerous hydrologic studies that monitor complex river systems using large data volumes of camera, video, or remotely sensed imagery.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2012
- Bibcode:
- 2012AGUFM.H31E1162G
- Keywords:
-
- 1827 HYDROLOGY / Glaciology;
- 1855 HYDROLOGY / Remote sensing;
- 1856 HYDROLOGY / River channels;
- 1926 INFORMATICS / Geospatial