Utilization of Object‐Oriented Software in the Image Analysis of Soil Thin Sections

Abstract

Computerized analysis of soil thin section imagery leads to quantitative estimation of various soil features and allows interpretation of soil properties in connection with soil genesis and evolution. Our objective was to assess an object‐oriented software (that was created especially for remote sensing) in the study of soil thin sections. The capabilities of this software were tested on three Orthic Humic Gleysols (Typic Humaquepts), profiles that have been thoroughly analyzed in the past by means of more traditional image analysis methods. The software utilizes a multiresolution segmentation function that enables the separation, at different levels of observation, of objects representing micromorphological features. Objects that corresponded to voids, coarse and fine fractions with variable compositions, and redoximorphic pedofeatures were distinguished and grouped in classes arranged in a hierarchical manner. Classification was based not only on object spectral values, but also on object shape characteristics, as well as class‐related features. Each designated class was evaluated quantitatively. Structural changes caused by agricultural practices were evidenced by image analysis in conjunction with micromorphological characterization. Relational and contextual data concerning Fe–Mn nodules and mottles were extracted and interpreted from the perspective of pedogenetic conditions.