Soil classification with spaceborne multi-temporal hyperspectral imagery using spectral unmixing and image fusion

Date

2023-3-24

Author

Kaba, Eylem

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

279
views

296
downloads

Cite This

Soil maps are important data sources for many agricultural or environmental studies. Satellites and airborne platforms carrying hyperspectral sensors provide new possibilities for the estimation of soil properties. However, the main obstacle in soil classification with remote sensing methods is the vegetation whose spectral signature mixes with that of the soil. The objective of this thesis is to detect soil texture properties after eliminating the effects of vegetation using hyperspectral imaging data. First, the endmembers common to all images and their abundances are estimated. Then the endmembers are classified as stable ones (soil, rock, etc.) and unstable ones (green vegetation, dry vegetation, etc.). The method eliminates vegetation from the images with orthogonal subspace projection and fuses multiple images with weighted mean for better signal-to-noise-ratio. Finally, the fused image is classified with the random forest technique to obtain the soil maps. The method is tested on synthetic and real images in an area in Texas, USA. With three synthetic images, individual classification results are 81.78%, 79.84%, and 86.33%. After OSP, the rates increase to 85.70%, 88.21%, and 91.78%, respectively, while it increases to 91.85% with fusion. With real images from the dates 22/06/2013, 25/09/2013, and 24/10/2013, the classification accuracies increase from 70.51%, 68.87%, and 63.18% to 71.96%, 71.78%, and 64.17%, respectively. Fusion provides a better improvement in classification with 75.27% accuracy. The results show that the method can improve classification accuracy with the elimination of vegetation contribution. The approach is promising and can be applied to various other classification tasks.

Subject Keywords

Soil classification, Hyperspectral, Random forest, Unmixing, Image fusion

URI

https://hdl.handle.net/11511/102864

Collections

Graduate School of Natural and Applied Sciences, Thesis