基于深度学习的林区多源数据超分辨率模型构建——以天山东部国有林管理局板房沟分局为例

Construction of super-resolution model of multi-source data based on deep learning-taking the Branch of Stateowned Forest Administration in eastern Tianshan Mountains as an example

以天山东部国有林管理局板房沟分局为例,建立了适用于森林经营单位尺度的多地物类型、多尺度、多源数据的超分辨率模型。文章构建的模型较其他模型在此研究区有更好的稳定性和抗干扰性,且在主观效果对比中,能够更全面地提取低分辨率图像特征,更充分地恢复图像的纹理信息,在一定程度上满足了林业生产的部分需求。

Taking the Branch of State-owned Forest Administration in eastern Tianshan Mountains as an example,a super-resolution model of multi-ground type,multi-scale and multi-source data suitable for forest management unit scale is established.The model constructed in this paper has better stability and anti-interference than other models in this study area,and in the subjective effect comparison,it can extract low-resolution image features more comprehensively,more fully recover the texture information of the image,and meet some of the needs of forestry production to a certain extent.