A vertical profile imaging method for quantifying rock fragments in gravelly soil

• Published in: Catena (Giessen) Vol. 193; p. 104590
• Main Authors: Jiang, Zhuo-Dong, Wang, Qiu-Bing, Adhikari, Kabindra, Brye, Kristofor R., Sun, Zhong-Xiu, Sun, Fu-Jun, Owens, Phillip R.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2020
• Subjects: Image segmentation; Profile imaging; Rock fragments; Soil profile; Profile imaging; Image segmentation; Rock fragments; Soil profile
• ISSN: 0341-8162; 1872-6887
• DOI: 10.1016/j.catena.2020.104590

•A profile imaging method was developed for estimating rock fragment concentrations.•Soil profile imaging method was compared with visual and direct measurement.•An alternative image-processing method compared well to other standard methods.•Profile imaging method may save time estimating soil rock fragment concentrations. Rock fragments (RFs) play an important role in soils that can influence land use and management decisions. Rock fragment concentrations most often are visually estimated (VE) in the field, for which accuracy strongly depends on experience and practice of soil classifiers. The direct measurement (DM) method is more accurate, but is costly and labor-intensive. Due to its importance, a simple and inexpensive method for estimating RF concentrations without destructive soil sampling is necessary. The development of camera technology and digital image processing provide an opportunity for soil descriptions and quantitative analyses using an alternative soil profile imaging (PI) method. Therefore, the objective of this study was to explore the potential for RF concentration estimation using a PI method as compared to the VE and DM methods. Apart from VE and DM, different image processing procedures inside the PI method were compared in a natural, gravelly, alluvial profile (Entisol) with low contrast among RFs in order to develop an alternative and efficient RF quantification method. Results showed that the hue-saturation-value (HSV) color model performed better than the red-greenblue color model for identifying RFs. The proposed profile imaging method accurately segmented 84.9% of the RFs from the soil matrix. The PI had a greater correlation with DM (r = 0.81) compared to the VE method (r = 0.78). Results showed that the alternative PI method could be used to accurately and efficiently estimate RF concentrations in a natural, gravelly soil profile with low contrast among RFs.