The grid DEM(digital elevation model) generation can be from any of a number of sources:for instance,analogue to digital conversion of contour maps followed by application of the TIN model,or direct elevation point mo...The grid DEM(digital elevation model) generation can be from any of a number of sources:for instance,analogue to digital conversion of contour maps followed by application of the TIN model,or direct elevation point modelling via digital photogrammetry applied to airborne images or satellite images.Currently,apart from the deployment of point-clouds from LiDAR data acquisition,the generally favoured approach refers to applications of digital photogrammetry.One of the most important steps in such deployment is the stereo matching process for conjugation point(pixel) establishment:very difficult in modelling any homogenous areas like water cover or forest canopied areas due to the lack of distinct spatial features.As a result,application of automated procedures is sure to generate erroneous elevation values.In this paper,we present and apply a method for improving the quality of stereo DEMs generated via utilization of an entropy texture filter.The filter was applied for extraction of homogenous areas before stereo matching so that a statistical texture filter could then be applied for removing anomalous evaluation values prior to interpolation and accuracy assessment via deployment of a spatial correlation technique.For exemplification,we used a stereo pair of ASTER 1B images.展开更多
Height map estimation from a single aerial image plays a crucial role in localization,mapping,and 3D object detection.Deep convolutional neural networks have been used to predict height information from single-view re...Height map estimation from a single aerial image plays a crucial role in localization,mapping,and 3D object detection.Deep convolutional neural networks have been used to predict height information from single-view remote sensing images,but these methods rely on large volumes of training data and often overlook geometric features present in orthographic images.To address these issues,this study proposes a gradient-based self-supervised learning network with momentum contrastive loss to extract geometric information from non-labeled images in the pretraining stage.Additionally,novel local implicit constraint layers are used at multiple decoding stages in the proposed supervised network to refine high-resolution features in height estimation.The structural-aware loss is also applied to improve the robustness of the network to positional shift and minor structural changes along the boundary area.Experimental evaluation on the ISPRS benchmark datasets shows that the proposed method outperforms other baseline networks,with minimum MAE and RMSE of 0.116 and 0.289 for the Vaihingen dataset and 0.077 and 0.481 for the Potsdam dataset,respectively.The proposed method also shows around threefold data efficiency improvements on the Potsdam dataset and domain generalization on the Enschede datasets.These results demonstrate the effectiveness of the proposed method in height map estimation from single-view remote sensing images.展开更多
基金Supported by the Ministry of Human Resource Development (MHRD),India (for Distinguished Institute Fellow)
文摘The grid DEM(digital elevation model) generation can be from any of a number of sources:for instance,analogue to digital conversion of contour maps followed by application of the TIN model,or direct elevation point modelling via digital photogrammetry applied to airborne images or satellite images.Currently,apart from the deployment of point-clouds from LiDAR data acquisition,the generally favoured approach refers to applications of digital photogrammetry.One of the most important steps in such deployment is the stereo matching process for conjugation point(pixel) establishment:very difficult in modelling any homogenous areas like water cover or forest canopied areas due to the lack of distinct spatial features.As a result,application of automated procedures is sure to generate erroneous elevation values.In this paper,we present and apply a method for improving the quality of stereo DEMs generated via utilization of an entropy texture filter.The filter was applied for extraction of homogenous areas before stereo matching so that a statistical texture filter could then be applied for removing anomalous evaluation values prior to interpolation and accuracy assessment via deployment of a spatial correlation technique.For exemplification,we used a stereo pair of ASTER 1B images.
基金supported by National Natural Science Foundation of China[grant number 42001329,42001283]Guangdong Basic and Applied Basic Research Foundation[grant number 2023A1515011718]+1 种基金China Postdoctoral Science Foundation[grant number 2021M701268]Foundation of Anhui Province Key Laboratory of Physical Geographic Environment,P.R.China[grant number 2022PGE012].
文摘Height map estimation from a single aerial image plays a crucial role in localization,mapping,and 3D object detection.Deep convolutional neural networks have been used to predict height information from single-view remote sensing images,but these methods rely on large volumes of training data and often overlook geometric features present in orthographic images.To address these issues,this study proposes a gradient-based self-supervised learning network with momentum contrastive loss to extract geometric information from non-labeled images in the pretraining stage.Additionally,novel local implicit constraint layers are used at multiple decoding stages in the proposed supervised network to refine high-resolution features in height estimation.The structural-aware loss is also applied to improve the robustness of the network to positional shift and minor structural changes along the boundary area.Experimental evaluation on the ISPRS benchmark datasets shows that the proposed method outperforms other baseline networks,with minimum MAE and RMSE of 0.116 and 0.289 for the Vaihingen dataset and 0.077 and 0.481 for the Potsdam dataset,respectively.The proposed method also shows around threefold data efficiency improvements on the Potsdam dataset and domain generalization on the Enschede datasets.These results demonstrate the effectiveness of the proposed method in height map estimation from single-view remote sensing images.
