Bank gully extraction from DEMs utilizing the geomorphologic features of a loess hilly area in China

As one of most active gully types in the Chinese Loess Plateau,bank gullies generally indicate soil loss and land degradation.This study addressed the lack of detailed,large scale monitoring of bank gullies and proposed a semi-automatic method for extracting bank gullies,given typical topographic features based on 5 m resolution DEMs.First,channel networks,including bank gullies,are extracted through an iterative channel bum-in algorithm.Second,gully heads are correctly positioned based on the spatial relationship between gully heads and their corresponding gully shoulder lines.Third,bank gullies are distinguished from other gullies using the newly proposed topographic measurement of "relative gully depth (RGD)."The experimental results from the loess hilly area of the Linjiajian watershed in the Chinese Loess Plateau show that the producer accuracy reaches 87.5%. The accuracy is affected by the DEM resolution and RGD parameters,as well as the accuracy of the gully shoulder line.The application in the Madigou watershed with a high DEM resolution validated the duplicability of this method in other areas.The overall performance shows that bank gullies can be extracted with acceptable accuracy over a large area,which provides essential information for research on soil erosion,geomorphology,and environmental ecology.

Height estimation from single aerial imagery using contrastive learning based multi-scale refinement network

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.