PSO-SLIC algorithm: A novel automated method for the generation of high-homogeneity slope units using DEM data

• Published in: Geomorphology (Amsterdam, Netherlands) Vol. 463; p. 109367
• Main Authors: Li, Yange, Fu, Bangjie, Han, Zheng, Fang, Zhenxiong, Chen, Ningsheng, Hu, Guisheng, Wang, Weidong, Chen, Guangqi
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.10.2024
• Subjects: DEM; High-homogeneity; PSO; SLIC; Slope units; Terrain segmentation; High-homogeneity; Slope units; SLIC; DEM; PSO; Terrain segmentation
• ISSN: 0169-555X
• DOI: 10.1016/j.geomorph.2024.109367

The generation of high-homogeneity slope units is crucial for terrain understanding and further analysis. Currently, the parameterization of the slope units extraction method largely depends on empirical adjustments, making the optimization process notably time-consuming. Hence, automated and efficient segmentation of slope units remains a major challenge. To address this, the PSO-SLIC algorithm, an innovative method that integrates the Simple Linear Iterative Clustering (SLIC) and Particle Swarm Optimizing (PSO) algorithms for slope unit generation, is proposed. Specifically, SLIC is adopted as the basic slope unit generation method, wherein the embedded PSO is designed to optimize the essential parameter combinations for SLIC. Additionally, a new fitness function for PSO is proposed to comprehensively consider the homogeneity, heterogeneity, and shape requirements of slope units during the generation process. Notably, the proposed PSO-SLIC algorithm is parameter-free and enables adaptive selection of the optimal parameters without empirical adjusting. In this study, two separate study areas are employed to validate the performance of the proposed method. Specifically, in site A, our method generated a lower percentage (5.16 %) of highly heterogeneous slope units, compared to the hydrological method (15.77 %) and the multi-resolution segmentation method (11.10 %). The statistical results in site B further validate the superiority of our proposed method. Moreover, the proposed PSO-SLIC method effectively reduces the required number of units for terrain representation and decreases subsequent computational demands. Results demonstrate that the proposed PSO-SLIC algorithm can be used for refined terrain segmentation reliably, providing a practical solution for terrain analysis and related applications. •A parameter-free slope units generation method is proposed.•The homogeneity, heterogeneity, and squareness is fully accounts in this method.•Our proposed method can be used for automated and efficient terrain segmentation.