Image Analysis and Classification Based on Soil Strength

Abstract

Satellite imagery classification is useful for a variety of commonly used applications, such as land use classification, agriculture, wetland delineation, forestry, geology, and landslide potential. However, image classification for physical properties of surface soils, such as strength or bearing capacity, is often obscured by other surface conditions, such as moisture and vegetation, although these are also indicators of soil strength. This project used remote methods of terrain analysis to search for areas suitable for vehicle or aircraft maneuverability based on slope, roughness, vegetation, soil type, and wetness and also performed direct classification of imagery based on soil strength. Using a maximum likelihood supervised classification approach, trained by a limited amount of ground-truth strength measurements, a soil strength classification was applied to WorldView-2 multispectral satellite imagery. This paper presents the work done on the imagery classification for soil strength, the apparent relationship between the reflectance and soil strength, and the ongoing work to expand the technique to new imagery by using existing training sets.

Terrestrial Geospatial Remote Assessment for Ingress Locations (Terrestrial GRAIL)

“…The findings of Sopher et al (2016bSopher et al ( , 2016c demonstrate that initial assessments of multispectral satellite imagery classifications and ground spectra evaluation show potential for classification related to terrain strength. In Sopher's study, eight strength categories were adequate for classifying an image for terrain strength.…”

“…The GRAIL team used data gathered at FHL to develop remote methods of terrain assessment to search for areas suitable for entry and maneuver based on slope, roughness, vegetation, soil type, and wetness (Shoop et al 2018b) and to classify imagery for soil strength (Sopher et al 2016b(Sopher et al , 2016c). Sopher used a maximum-likelihood supervised classification trained by a limited number of ground truth measurements or user input to apply a soil strength classification to WV2 multispectral satellite imagery, achieving encouraging results.…”

“…The team also used soil moisture and strength measurements to validate and refine the CBR soil strength model. Sopher et al (2016bSopher et al ( , 2016c and West (2017) discuss in greater detail the concept of field verification and image feature detection work, respectively. Shoop et al (2018b) and Sinclair et al (2017Sinclair et al ( , 2018 present verification, refinement, and further validation.…”

“…While this is counterintuitive because strength or stiffness are related to a physical loading of the surface, the bearing capacity of the terrain can be related to other parameters that are expressed at the surface, such as the type, quantity, quality of the vegetation, soil texture and mineralogy, or other parameters that we still are not able to broadly classify, such as grain angularity. Sopher et al (2016bSopher et al ( , 2016c explored image classification for strength. Classification is the process of categorizing each pixel of an image into a certain spectral class to describe what kind of object-or in the case of GRAIL, land cover, obstacle, or soil strength-that pixel depicts.…”

Soil strength analysis of Sonoran Desert landforms

“…They developed multiple approaches using basic terrain data, including remotely sensed data and imagery, to predict detailed surfaces and subsurface terrain conditions over a wide range of scales in near real time. Other recent efforts have also used remotely sensed data to create geomorphic maps that predict dust emission potential at scales ranging from individual deserts to continents (Parajuli and Zender 2017;Bullard et al 2011) and using remotely sensed data to infer soil strength (Sopher et al 2016a(Sopher et al , 2016b).…”