Effective Identification of Terrain Positions from Gridded DEM Data Using Multimodal Classification Integration

ISPRS International Journal of Geo-Information

Mingwei Zhao¹  Ling Jiang²  Qiuhua Liang²  Chun Wang²  Kai Liu³  Dequan Ling⁴ 

[1] Technology, Nanjing 210044, China;Anhui Engineering Laboratory of Geo-information Smart Sensing and Services, Chuzhou University, Chuzhou 239000, China;School of Architecture, Building and Civil Engineering, Loughborough University, Leicestershire LE11 3TU, UK;;School of Geography and Remote Sensing, Nanjing University of Information Science &

关键词: terrain position;    DEM;    multimodal classification;    segmentation;    digital terrain analysis;   

DOI  :  10.3390/ijgi7110443

来源: DOAJ

【 摘 要 】

Terrain positions are widely used to describe the Earth’s topographic features and play an important role in the studies of landform evolution, soil erosion and hydrological modeling. This work develops a new multimodal classification system with enhanced classification performance by integrating different approaches for terrain position identification. The adopted classification approaches include local terrain attribute (LA)-based and regional terrain attribute (RA)-based, rule-based and supervised, and pixel-based and object-oriented methods. Firstly, a double-level definition scheme is presented for terrain positions. Then, utilizing a hierarchical framework, a multimodal approach is developed by integrating different classification techniques. Finally, an assessment method is established to evaluate the new classification system from different aspects. The experimental results, obtained at a Loess Plateau region in northern China on a 5 m digital elevation model (DEM), show reasonably positional relationship, and larger inter-class and smaller intra-class variances. This indicates that identified terrain positions are consistent with the actual topography from both overall and local perspectives, and have relatively good integrity and rationality. This study demonstrates that the current multimodal classification system, developed by taking advantage of various classification methods, can reflect the geographic meanings and topographic features of terrain positions from different levels.

【 授权许可】

Unknown