Influence of GLCM texture parameters on lithological mapping using Sentinel-1 imagery

Abstract

Gray Level Co-Occurrence Matrix (GLCM) textures demonstrate great potential in lithological mapping, yet the influence of GLCM parameters (window size, distance, and angle) on mapping lithology using Sentinel-1 images has never been thoroughly explored. Textures calculated with different GLCM parameters using Sentinel-1 data under different incidence angles were used for lithological classification in a semi-arid area. We found large incidence angle can improve the classification result slightly; larger window size and smaller distance with an appropriate angle can lead to a higher classification accuracy; different parameter combinations can result in a maximum difference in classification accuracy of about 10%, and the highest accuracy (76.27%) is achieved by using a window size of 15*15 with a distance of 1 at the angle of 0°. Our study enhances the application of Sentinel-1-derived GLCM textures in geological fields, the promising classification accuracy indicates more opportunity for lithology discrimination using SAR data compared to earlier studies.