Geographically Weighted Spatial Unmixing for Spatiotemporal Fusion

Abstract

Spatio-temporal fusion is a technique applied to create images with both fine spatial and temporal resolutions by blending images with different spatial and temporal resolutions. Spatial unmixing is a widely used approach for spatio-temporal fusion, which requires only the minimum number of input images. However, ignorance of spatial variation in land cover between pixels is a common issue in existing spatial unmixing methods. For example, all coarse neighbors in a local window are treated equally in the unmixing model, which is inappropriate. Moreover, determination of the appropriate number of clusters in the known fine spatial resolution image remains a challenge. In this paper, a geographically weighted spatial unmixing (SU-GW) method was proposed to address the spatial variation in land cover and increase the accuracy of spatio-temporal fusion. SU-GW is a general model suitable for any spatial unmixing methods. Specifically, the existing regularized version and soft classification-based version were extended with the proposed geographically weighted scheme, producing 24 versions (i.e., 12 existing versions were extended to 12 corresponding geographically weighted versions) for spatial unmixing. Furthermore, the cluster validity index of Xie and Beni (XB) was introduced to determine automatically the number of clusters. A systematic comparison between the experimental results of the 24 versions indicated that SU-GW was effective in increasing the prediction accuracy. Importantly, all 12 existing methods were enhanced by integrating the SU-GW scheme. Moreover, the identified most accurate SU-GW enhanced version was demonstrated to outperform two prevailing spatio-temporal fusion approaches in a benchmark comparison. Therefore, it can be concluded that SU-GW provides a general solution for enhancing spatio-temporal fusion, which can be used to update existing methods as well as future potential versions.

Pansharpening and spatiotemporal image fusion method for remote sensing

“…Strong temporal shifts are difficult to forecast, but they can be detected via methods such as spatial filtering. One popular method for fusing spatial and temporal information is spatial unmixing (SU) [137]. Nevertheless, the diverse range of landscapes necessitates that distinct neighboring coarse pixels make varying contributions to the core pixel.…”

“…Retain spatial and spectral properties while integrating temporal traces Large computational cost, less computation efficiency compared to STARFM and Fit-FC FSDAF [161] Provide a fused image in a relatively swift manner, fast running speed Compromised performance IFSDAF [162] Improve the resolution of images Seamless spatial prediction results, leading to the loss of spatial details SFSDAF [163] Improved prediction accuracy Computational burden, uncertainty in prediction model geographical weighting (GW)-based SU technique [137] is used; however, using nearby pixels for prediction may cause image blurring and the loss of high-frequency features.…”

An Experimental Study of the Accuracy and Change Detection Potential of Blending Time Series Remote Sensing Images with Spatiotemporal Fusion

“…Following the framework proposed by Zhukov et al [24], unmixing-based methods [25][26][27][28] employ spatial unmixing techniques for fusion, which estimate the highresolution endmembers by unmixing the coarse-resolution pixels using the class scores explained by the reference image. Due to the wide spectrum and large resolution ratio, the unmixing-based methods may be prone to errors in abundance estimation, spectral variations, and nonlinear mixing.…”

“…Its revisions, such as SFSDAF [38], FSDAF 2.0 [39], and EFSDAF [40], can also be categorized into this type. Other hybrid studies [28,41] integrated weight and unmixing strategies, too. Our previous work [42] is also a hybrid type, which integrates the results of FSDAF and Fit-FC to enhance performance.…”

Spatial sampling, data models, spatial scale and ontologies: Interpreting spatial statistics and machine learning applied to satellite optical remote sensing

“…Lying between measurement and the ability to identify and label the original objects of interest is the concept of the 'mixed pixel' (Peng et al, 2022). Mixed pixels occur when more than one object class contributes to the overall signal measured and allocated to a pixel.…”