A framework for extracting and characterizing landform structures under buildings using elevation thresholds and K-means algorithm

The study of landform characterization is important in geography and environmental science, where complex landforms and buildings pose challenges. Accurate landform characterization is crucial for urban planning and environmental modelling. Unfortunately, the analysis of uneven landforms particularly under buildings is understudied, affecting the quality of integration between digital terrain and high level-of-detail building models. To address this, a framework is presented in this study that combines elevation thresholding filters and the k-means clustering algorithm to capture subtle landform varieties that may affect the integration between digital terrain and high level-of-detail building models. This approach integrates spatial and elevation data, capturing subtle changes that conventional methods may miss. The framework uses adaptive threshold values based on cell elevation standard deviation to create different elevation classes. Within each class, the k-means algorithm clusters vector points based on spatial and elevation attributes. The framework was evaluated using different techniques, demonstrating its effectiveness. Notably, it is effective at identifying buildings constructed on uneven slopes, particularly those under the ½ σ threshold. The proposed framework enhances the accuracy of integrating high-detail building models.