gabor: Rock glacier remote sensing using Gabor texture filters

In alexanderbrenning/wiml: Interpreting machine-learning models in transformed space

Rock glacier remote sensing using Gabor texture filters

Description

This data set for binary classification tasks contains information on the presence/absence of flow structures related to the deformation of rock glaciers in the Andes of Santiago, and corresponding remotely-sensed texture attributes and terrain attributes as predictors. Rock glaciers are ice-debris landforms - the visible expression of creeping mountain permafrost.

Usage

gabor

Format

A data frame with 4656 grid cells (1290 from flow patterns and 3366 from other terrain outside of rock glaciers) in two study areas (area = "LN": 2499 grid cells; area = "CAT": 2157 cells). (The flow patterns are from approximately 50 individual rock glaciers of varying size.)

class

Factor variable (levels: "TRUE", "FALSE") representing the presence ("TRUE") and absence ("FALSE") of rock glacier flow patterns.

x,y

UTM x and y coordinates (not to be used as predictors)

area

Factor variable identifying one of the two study areas, "LN" or "CAT".

dem

Elevation in metres above sea level (m a.s.l.)

slope

Local slope angle in degrees

cslope

Slope angle of the upslope contributing area in degrees

log.carea

Logarithm (to the base 10) of the size of the upslope contributing area in m²

log.cheight

Logarithm (to the base 10) of the height of the upslope contributing area in m

pisr

Annual potential incoming solar radiation

m30e1g5rg,m30e1g5max,m30e1g5min,m30e1g5med,m30e1g10rg,m30e1g10max,m30e1g10min,m30e1g10med,m30e1g20rg,m30e1g20max,m30e1g20min,m30e1g20med,m30e1g30rg,m30e1g30max,m30e1g30min,m30e1g30med,m30e1g50rg,m30e1g50max,m30e1g50min,m30e1g50med,m30e2g5rg,m30e2g5max,m30e2g5min,m30e2g5med,m30e2g10rg,m30e2g10max,m30e2g10min,m30e2g10med,m30e2g20rg,m30e2g20max,m30e2g20min,m30e2g20med,m30e2g30rg,m30e2g30max,m30e2g30min,m30e2g30med,m30e2g50rg,m30e2g50max,m30e2g50min,m30e2g50med

Gabor features of the form m30e*i*g*j*x with the following settings (see Brenning et al. 2012 for details): i = axis ratio (1 or 2) of Gabor filter; j = wavelength of Gabor filter (5, 10, 20, 30, or 50 m); x = aggregation scheme (“min” = minimum; “max” = maximum; “rg” = range; “med” = median

Details

Texture attributes derived from high-resolution panchromatic IKONOS imagery form the largest group of features in this study, and terrain attributes are used as additional predictors. A ‘filter bank’ of Gabor filters is used since Gabor features are capable of detecting ‘zebra stripe’ type patterns that relate to the troughs and ridges typically found on rock glaciers. The second group of features represent topographic conditions, e.g. elevation, slope angle and size of the upslope contributing area. These are proxies for topoclimatic conditions that may relate to the formation and conservation of permafrost, and for talus supply to these ice-debris landforms.

This data set is a subset of the data used by Brenning et al. (2012), specifically a subset of the Laguna Negra area. Note that areas that can “obviously” not present rock glaciers have been masked out (i.e. removed from the data set), e.g. steep slopes, in order to allow the classifier to focus on the “difficult” areas; see Brenning et al. (2012) for details.

Objective: To identify rock-glacier flow patterns based on the available texture and terrain attribute data.

Source

Brenning, A., Long, S. and Fieguth, P. 2012. Detecting rock glacier flow structures using Gabor filters and IKONOS imagery. Remote Sensing of Environment, 125: 227-237. https://doi.org/10.1016/j.rse.2012.07.005

alexanderbrenning/wiml documentation built on Sept. 29, 2023, 4:45 a.m.