Facing the 3rd national land survey (cultivated land quality): soil survey application for soil texture detection based on the high-definition field soil images by using perceptual hashing algorithm (pHash)

• Published in: Journal of soils and sediments 2020-09, Vol.20 (9), p.3427-3441
• Main Authors: Pan, Hanyue, Liang, Jia, Zhao, Ye, Li, Fangfang
• Format: Article
• Language: English
• Subjects: Algorithms; Arable land; Composition; Cultivated lands; Detection; Earth and Environmental Science; Environment; Environmental Physics; Field investigations; Geographical locations; Hash based algorithms; High definition; Image processing; Image resolution; Laboratories; Land surveys; Machine learning; Measurement; Pipettes; Pretreatment; Sec 5 • Soil and Landscape Ecology • Research Article; Soil; Soil analysis; Soil investigations; Soil properties; Soil Science & Conservation; Soil surveys; Soil texture; Soils; Statistical analysis; Statistical methods; Surveying; Surveys; Texture; Training; Verification
• ISSN: 1439-0108; 1614-7480
• DOI: 10.1007/s11368-020-02657-5

Purpose Land and soil surveys are of significance to investigate the conditions of land or soil in certain regions. The determination of soil texture type relies on the fundamental laboratory measurement of soil physical granulometric composition. Traditional approaches cannot be used finely in the field. In this paper, the relationship between soil texture and high-resolution field soil images was established. The android application was developed based on the pHash to determine soil texture type in the field investigation. Materials and methods The whole experiment is divided into three sections—preparation of field and standardized soil samples, development of the android application, and the verification of the android application. A total of 37 arable soil samples from different geographical locations with various typical soil texture types were taken during the national land survey in China. Necessary pretreatment with soil samples was done before laboratory analysis. The standardized soil image database was established by uploading standardized soil images. The JAVA language is applied to realize the pHash of the application by IntelliJ IDEA. Physical particle clay and sand content are the chosen indicators to describe the soil granulometric composition quantificationally for the verification part—these two contents of each sample were measured by both the pipette method and the android application more than three times separately—then contrast the testing results to indicate the performance of the android application. Results and discussion Fitting performances of the pipette method and android application reached 89.23% in all group results. The statistical analysis of the difference between the two approaches is not significant ( p  > 0.05). It is believed that increasing the training number can improve this android application in subsequent studies. Our research changes the idea of determining the soil texture from direct measurements to intermediate comparison, which makes soil texture in-field detection feasible. Conclusions This android application extends the measure tools by learning the thought of computerized algorithms. The measurement results of the application show accuracy and repeatability during the determination of soil granulometric composition, compared with the ones of the pipette method. Android application based on the perceptual hashing algorithm can be friendly used during land and soil surveys, as well as other field studie