Matlab Software for Supervised Habitat Mapping of Freshwater Systems Using Image Processing
Open Access
- 3 December 2021
- journal article
- research article
- Published by MDPI AG in Remote Sensing
- Vol. 13 (23) , 4906
Abstract
We present a software package for the supervised classification of images useful for cover-type mapping of freshwater habitat (e.g., water surface, gravel bars, vegetation). The software allows the user to select a representative subset of pixels within a specific area of interest in the image that the user has identified as a cover-type habitat of interest. We developed a graphical user interface (GUI) that allows the user to select single pixels using a dot, line, or group of pixels within a defined polygon that appears to the user to have a spectral similarity. Histogram plots for each band of the selected ground-truth subset aid the user in determining whether to accept or reject it as input data for the classification processes. A statistical model, or classifier, is then built using this pixel subset to assign every pixel in the image to a best-fit group based on reflectance or spectral similarity. Ideally, a classifier incorporates both spectral and spatial information. In our software, we implement quadratic discriminant analysis (QDA) for spectral classification and choose three spatial methods—mode filtering, probability label relaxation, and Markov random fields—to incorporate spatial context after computation of the spectral type. This multi-step interactive process makes the software quantitatively robust, broadly applicable, and easily usable for cover-type mapping of rivers, their floodplains, wetlands often components of these functionally linked freshwater systems. Indeed, this supervised classification approach is helpful for a wide range of cover-type mapping applications in freshwater systems but also estuarine and coastal systems as well. However, it can also aid many other applications, specifically for automatic and quantitative extraction of pixels that represent the water surface area of rivers and floodplains.Keywords
This publication has 24 references indexed in Scilit:
- A multi-model ensemble approach to seabed mappingJournal of Sea Research, 2015
- River flood plains are model ecosystems to test general hydrogeomorphic and ecological conceptsRiver Research and Applications, 2010
- Remote Sensing of Channels and Riparian Zones with a Narrow-Beam Aquatic-Terrestrial LIDARRemote Sensing, 2009
- The effect of spatial autocorrelation and class proportion on the accuracy measures from different sampling designsISPRS Journal of Photogrammetry and Remote Sensing, 2008
- Classifier Technology and the Illusion of ProgressStatistical Science, 2006
- USING AIRBORNE MULTISPECTRAL IMAGERY TO EVALUATE GEOMORPHIC WORK ACROSS FLOODPLAINS OF GRAVEL‐BED RIVERSEcological Applications, 2005
- An assessment of the effectiveness of decision tree methods for land cover classificationRemote Sensing of Environment, 2003
- Riverine flood plains: present state and future trendsEnvironmental Conservation, 2002
- An evaluation of some factors affecting the accuracy of classification by an artificial neural networkInternational Journal of Remote Sensing, 1997
- Extensions of linear discriminant analysis for statistical classification of remotely sensed satellite imageryMathematical Geology, 1980