An Autonomous Program for Crack Length Calculation in an Unsaturated Soil in 1-D Column

Abstract

Cracks are widely present in natural and engineered soils. Their presence has the potential to significantly alter both hydraulic and mechanical properties of soil, thereby causing the performance to weaken with respect to different engineering disciplines, particularly geotechnical, environmental and geological engineering. This makes the quantification and characterization of cracks important, which would lead to a better understanding about their extent and complexity in soil under various conditions of seasonal rainfall changes. The length of individual cracks has been studied in the literature to determine the smallest volume with respect to which an averaging of a certain property of the cracked soil can be performed. In this regard, the analysis of images has been instrumental in crack length quantification. However, the currently used software-based methods involving manual adjustment through visual inspection of the cracks in two-dimensional surface images are a time-consuming task that lacks accuracy, sensitivity and reproducibility. There is, hence, the need to develop automated imaging techniques for analyzing crack length in drying soils rather than relying on software-based methods. The main objective of this study is to introduce a novel image analysis tool that employs an open source computer vision library coded in Python to effectively characterize crack length in soils. A simple experimental setup was developed using a one-dimensional column containing compacted red soil in an environment-controlled chamber. A series of images of soil sample was captured using commercially available camera model (Canon EOS 700D) to have a photographic time-lapse representation of the cracking process. A step-by-step strategy using a Python script is presented here to outline an approach to divide the circular two-dimensional image into equal radial slices and calculate the crack length in each slice. The approach can be further extended to calculate other crack parameters sector-wise in a circular image.