遥感图像超分辨率重建技术研究及地质解译应用  

Study on super-resolution reconstruction technique of remote sensing image and application of geological interpretation

作　　者：万昫宏 赵英俊[1] WAN Xuhong;ZHAO Yingjun(National Key Laboratory of Remote Sensing Information and Image Analysis Technology Beijing Research Institute of Uranium Geology,Beijing 100029,China)

机构地区：[1]核工业北京地质研究院遥感信息与图像分析技术国家级重点实验室,北京100029

出　　处：《世界核地质科学》2024年第3期584-594,共11页World Nuclear Geoscience

摘　　要：遥感图像超分辨率重建技术通过提升遥感图像的空间分辨率,增强图像的细节和清晰度,在地质调查中能够提高地质特征的识别精度和信息提取的准确性,具有重要的应用价值。针对现有遥感图像重建模型在地质解译下的细节和特征信息丢失等缺点,提出一种基于增强型深度残差网络的多尺度渐进式残差网络(MPEDSR)的超分辨率重建模型。该模型重点针对残差结构进行调整,在残差块堆叠结构中加入跳跃连接,同时在残差块间引入通道注意力机制模块,实现残差分支信息的充分利用,并通过测试集图像证明所提方法的地质调查要素解译能力。实验结果表明:所提算法在主观视觉感受及峰值信噪比(PeakSignal-to-NoiseRatio,PSNR)和结构相似度(Structural Similarity,SSIM)等客观评价指标上均高于现有几种主流算法。在2倍和4倍放大因子下相较原模型在PSNR上分别提升0.176和0.194 dB,在SSIM上分别提升0.018和0.021。此外,该网络能有效提取滑坡和岩性边界,识别断裂构造。本研究为高分辨率遥感图像下的地质解译和地质灾害监测提供了有效的技术手段,推动了遥感地质解译的精细化和智能化发展。The super resolution reconstruction technology of remote sensing image can improve the spatial resolution of remote sensing image,enhance the detail and clarity of the image,and improve the accuracy of feature identification and information extraction in geological investigation.To reduce the loss of detail and feature information by the existing reconstruction models of remote sensing image for geology application,a multi-scale progressive enhanced deep residual super-resolution network(MPEDSR)was proposed on the enhanced deep super-resolution network.This model focuses on adjusting the residuals structure,adding skip connections to the residual stack structure,and introducing channel attention mechanism module between the residuals to make full use of the residuals branch information.The test data proves the interpretation ability of geological survey elements of the proposed method.The experimental results showed that the proposed algorithm was higher in subjective visual perception,Peak Signal-to-Noise Ratio(PSNR),Structural Similarity(SSIM)and other objective evaluation indexes than the existing algorithms.Compared with the original model,the PSNR was improved by 0.176 dB and 0.194 dB at 2x and 4x scale factors respectively.SSIM increased by 0.018 and 0.021 respectively.In addition,the network can effectively extract landslide and lithology boundaries and identify fault structures.Therefore it provides an effective technical means to improve the high-resolution remote sensing images for geological interpretation and geological hazard monitoring,and promotes the refinement and intelligent development of remote sensing geological interpretation.

关 键 词：超分辨率重建 残差网络 遥感图像 地质解译 

分 类 号：P23[天文地球—摄影测量与遥感] TP391[天文地球—测绘科学与技术]