Application of transfer learning of deep CNN model for classification of time-series satellite images to assess the long-term impacts of coal mining activities on land-use patterns

• DOI: 10.1080/10106049.2022.2057595
• Publication Date: 2022-03-25T12:17:18Z
• Authors: Ajay Kumar; Amit Kumar Gorai
• Keywords: Land Cover

The study aims to analyse the long-term impacts of mining activities in Jharia coalfield (JCF) on land-use (LU) patterns using transfer learning deep convolutional neural network (Deep CNN) model. A new database was prepared by extracting 10,000 image samples 6 × size for five LU types (barren land, built-up area, coal region, vegetation and waterbody) from Landsat data train validate satellite 1987 2021 at an interval two years used change analysis. results revealed that model offers 95 88% accuracy training validation dataset. indicate barren waterbody have been decreased 237.30 sq. km. (=39.88%) 171.25 km (=28.78%), 118.77 (=19.96%) 68.73 (=11.55%), 35.58 (=5.98%) 18.68 (=3.14%) during 1987–2021, respectively. On other hand, area increased 120.14 (=20.19%) 233.02 (=39.16%) 83.19 (=13.98%) 103.36 (=17.37%) 1987–2021. time-series correlation is most sensitive type 2021, whereas land least up 2011, thereafter sensitive.