Image Recognition for Large Soil Maps Archive Overview: Metadata Extraction and Georeferencing Tool Development

• Published in: Data Analytics and Management in Data Intensive Domains Vol. 1620; pp. 193 - 204
• Main Authors: Vasilyeva, Nadezda A., Vladimirov, Artem, Vasiliev, Taras
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2022; Springer International Publishing
• Series: Communications in Computer and Information Science
• Subjects: Archive collection; Heterogeneous data; Image recognition; Reproducibility; Soil maps
• ISBN: 9783031122842; 3031122844
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-3-031-12285-9_12

During the second half of 20th century Dokuchaev Soil Science Institute has collected soil maps for a half of the Eurasian continent as a result of large national soil surveys which lasted for several decades with the efforts of the former USSR. Such labor-intensive expeditions on countries scale were not repeated since then. The question of future soil dynamics as Earth’s fertile layer became crucial with global population growth and causes large part of uncertainty in Earth System Modelling. Most of the present knowledge about soil types is still in form of paper soil maps, representing valuable knowledge about soil cover of the past. Soil type itself is a crucial factor which still cannot be determined remotely but can be updated. Archive soil maps (several thousands of sheets) are an example of data which require digitizing and could profit from application of image recognition techniques. In the current study we present a demo tool for fast extraction of metadata and geo-referencing of paper soil maps using image recognition techniques. Presented software can be used for creating soil maps digital catalog allowing for a quick overview of a large collection.