Texture–terrain integration for classification of karst Fenglin–Fengcong landforms in Southwest China

The classification of Fenglin–Fengcong karst landforms is complicated by morphological heterogeneity, spectral similarity and indistinct landform boundaries. This study classifies Fenglin–Fengcong landforms in Southwest China by integrating high‐resolution GF‐1 multispectral imagery with ALOS PALSAR digital elevation model (DEM) data. Spectral bands, vegetation indices, terrain derivatives and image texture features were extracted to support landform classification. Two texture descriptors, the grey‐level co‐occurrence matrix (GLCM) and local binary pattern (LBP), were incorporated into an object‐based Random Forest (RF) classification framework. The analysis was conducted in three representative karst regions—Guilin, Luoping, and Xingyi—to examine texture performance under contrasting geomorphic conditions. The results showed that terrain derivatives improved the delineation of karst morphological structure compared with spectral features alone, while image texture features further enhanced boundary recognition and reduced confusion between karst landforms and surrounding vegetation. Differences in texture performance were observed among regions with contrasting landform structures, with LBP showing more stable performance in areas characterised by dense peak clustering and pronounced structural heterogeneity. The highest classification accuracy was achieved using the LBP‐based texture model (overall accuracy = 92.13%, F ‐score = 95.96%). These results indicate that the contribution of texture features to karst landform classification is closely related to geomorphic structure and spatial scale, highlighting the importance of considering regional geomorphic characteristics when applying texture‐based remote sensing methods to karst landscapes.