.png){kind=icon}
返回
A spatial interpolation method based on denoising diffusion probabilistic model
.png){kind=icon}
PRE
AI
DOI：10.1080/13658816.2025.2512219.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Spatial interpolation is critical in geographic information systems (GIS) and environmental science, particularly when dealing with high-dimensional and nonlinear data. Classical methods like Kriging and inverse distance weighting (IDW) often struggle with the complexities of irregular terrain and sparse datasets, and are inadequate for capturing the nonlinear characteristics of high-dimensional spatial data. In this paper, we introduce a novel interpolation method based on the Denoising Diffusion Probabilistic Model (DDPM), which incorporates ConvNeXt V2 blocks within a UNet architecture. To validate the performance of our model, we employ the Copernicus Digital Elevation Model (COP-DEM) dataset for simulation experiments. Experimental results demonstrate that the proposed DDPM method significantly outperforms classical interpolation techniques, particularly in scenarios with high-density control points, producing high-quality interpolation results with strong transferability. This approach shows considerable promise for spatial interpolation in high-dimensional, complex terrains, offering a more robust alternative to traditional methods. It not only addresses key challenges in interpolation accuracy but also opens up new possibilities for applying generative models in other spatial data processing domains, including environmental monitoring and geospatial modeling.
Keyword:
Spatial interpolation
denoising diffusion probabilistic model
convolutional neural network
deep learning
期刊
IF:
5.1
论文数:
2.7K
被引数:
9.3K
机构
引用论文
.png){kind=icon}
Kriging and Splines: An Empirical Comparison of their Predictive Performance in Some Applications
PREAI
分享收藏
查看更多内容