An object-based spatiotemporal fusion model for remote sensing images
An object-based spatiotemporal fusion model for remote sensing images
复制标题
DOI：
10.1080/22797254.2021.1879683
复制
发表时间：
2021-01
影响因子：
4
通讯作者：
Hua Zhang;Yue Sun;W. Shi;Dizhou Guo;Nanshan Zheng
中科院分区：
文献类型：
--
作者：
Hua Zhang;Yue Sun;W. Shi;Dizhou Guo;Nanshan Zheng
时空融合技术可以联合收割机综合不同影像的时间分辨率和空间分辨率的优点，实现对地球表面的连续监测，是解决遥感影像时空分辨率折衷的一种可行方案。在本文中，基于对象的时空融合模型（OBSTFM）提出了产生时空一致的数据，特别是在经历非形状变化（包括物候变化和土地覆盖变化没有形状变化）的地区。考虑到不同区域可能发生的不同变化，首先采用多分辨率分割来产生分割对象，然后引入线性注入模型来产生初步预测。此外，一个新的优化策略来选择相似的像素，以获得更准确的预测。利用两个非均匀区域的物候变化和土地覆盖类型变化的遥感数据集对所提方法进行了验证，实验结果表明，所提方法在非形状变化区域具有优势，在融合大尺度土地覆盖突变时具有令人满意的鲁棒性和可靠性。因此，OBSTFM具有监测高度动态景观的巨大潜力。
ABSTRACT Spatiotemporal fusion technique can combine the advantages of temporal resolution and spatial resolution of different images to achieve continuous monitoring for the Earth’s surface, which is a feasible solution to resolve the trade-off between the temporal and spatial resolutions of remote sensing images. In this paper, an object-based spatiotemporal fusion model (OBSTFM) is proposed to produce spatiotemporally consistent data, especially in areas experiencing non-shape changes (including phenology changes and land cover changes without shape changes). Considering different changes that might occur in different regions, multi-resolution segmentation is first employed to produce segmented objects, and then a linear injection model is introduced to produce preliminary prediction. In addition, a new optimized strategy to select similar pixels is developed to obtain a more accurate prediction. The performance of proposed OBSTFM is validated using two remotely sensed dataset experiencing phenology changes in the heterogeneous area and land cover type changes, experimental results show that the proposed method is advantageous in such areas with non-shape changes, and has satisfactory robustness and reliability in blending large-scale abrupt land cover changes. Consequently, OBSTFM has great potential for monitoring highly dynamic landscapes.