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.

Mapping of decentralised photovoltaic and solar thermal systems by remote sensing aerial imagery and deep machine learning for statistic generation

As a mean to monitor the rapid expansion of the highly decentralized PV market,identifying solar energy systems in aerial imagery by deep machine learning,is a research field that is getting increasing interest.One general challenge in the field is to create testing data of high quality that are representative of the end-use application.In this study we use the open source convolutional neural network developed within the DeepSolar project and apply it in the country of Sweden,for the purpose of generating market statistics,by scanning three complete municipalities for small decentralized photovoltaic and solar thermal systems.The evaluation of the performance is done against a highly accurate ground truth,which was created by cross-checking the classification results with the inventory of the local distribution system operators and the database of photovoltaic systems that have received a capital subsidy in Sweden,and combining that with physical onsite inspections.A process of generate additional training data and re-training the algorithm after each municipality scan was developed,which suc-cessively improved the accuracy,resulting in that 95%of all detectable photovoltaic,excluding building inte-grated and vertical systems,and 80%of all detectable solar thermal systems were correctly identified in the last municipality scan.The accurate ground truth allowed a quantification of why some systems are not detected.The generated dataset of solar energy systems could be connected to existing building and property inventories,which allowed creation of market segment statistics with remarkably high detail information.

Aerial-BiSeNet:A real-time semantic segmentation network for high resolution aerial imagery

The aircraft system has recently gained its reputation as a reliable and efficient tool for sensing and parsing aerial scenes.However,accurate and fast semantic segmentation of highresolution aerial images for remote sensing applications is still facing three challenges:the requirements for limited processing resources and low-latency operations based on aerial platforms,the balance between high accuracy and real-time efficiency for model performance,and the confusing objects with large intra-class variations and small inter-class differences in high-resolution aerial images.To address these issues,a lightweight and dual-path deep convolutional architecture,namely Aerial Bilateral Segmentation Network(Aerial-Bi Se Net),is proposed to perform realtime segmentation on high-resolution aerial images with favorable accuracy.Specifically,inspired by the receptive field concept in human visual systems,Receptive Field Module(RFM)is proposed to encode rich multi-scale contextual information.Based on channel attention mechanism,two novel modules,called Feature Attention Module(FAM)and Channel Attention based Feature Fusion Module(CAFFM)respectively,are proposed to refine and combine features effectively to boost the model performance.Aerial-Bi Se Net is evaluated on the Potsdam and Vaihingen datasets,where leading performance is reported compared with other state-of-the-art models,in terms of both accuracy and efficiency.

Rapid and robust slope failure appraisal using aerial photogrammetry and 3D slope stability models

Slope failures are an inevitable aspect of economic pit slope designs in the mining industry.Large open pit guidelines and industry standards accept up to 30%of benches in open pits to collapse provided that they are controlled and that no personnel are at risk.Rigorous ground control measures including real time monitoring systems at TARP(trigger-action-response-plan)protocols are widely utilized to prevent personnel from being exposed to slope failure risks.Technology and computing capability are rapidly evolving.Aerial photogrammetry techniques using UAV(unmanned aerial vehicle)enable geotechnical engineers and engineering geologists to work faster and more safely by removing themselves from potential line-of-fire near unstable slopes.Slope stability modelling software using limit equilibrium(LE)and finite element(FE)methods in three dimensions(3D)is also becoming more accessible,user-friendly and faster to operate.These key components enable geotechnical engineers to undertake site investigations,develop geotechnical models and assess slope stability faster and in more detail with less exposure to fall of ground hazards in the field.This paper describes the rapid and robust process utilized at BHP Limited for appraising a slope failure at an iron ore mine site in the Pilbara region of Western Australia using a combination of UAV photogrammetry and 3D slope stability models in less than a shift(i.e.less than 12 h).

No treeline shift despite climate change over the last 70 years

Background:The recent rise in temperature and shifting precipitation regimes threaten ecosystems around the globe to different degrees.Treelines are expected to respond to climate warming by shifting to higher elevations,but it is unclear whether they can track temperature changes.Here,we integrated high-resolution aerial imagery with local climatic and topographic characteristics to study the treeline dynamic from 1945 to 2015 on the semiarid Mediterranean island of Crete,Greece.Results:During the study period,the mean annual temperature at the treeline increased by 0.81℃,while the average precipitation decreased by 170 mm.The treeline is characterized by a diffuse form,with trees growing on steep limestone slopes(>50°)and shallow soils.Moreover,the treeline elevation decreases with increasing distance from the coast and with aspect(south>north).Yet,we found no shift in the treeline over the past 70 years,despite an increase in temperature in all four study sites.However,the treeline elevation correlated strongly with topographic exposure to wind(R^(2)=0.74,p<0.001).Therefore,the temporal lag in treeline response to warming could be explained by a combination of topographic and microclimatic factors,such as the absence of a shelter effect and a decrease in moisture.Conclusion:Although there was no treeline shift over the last 70 years,climate change has already started shifting the treeline altitudinal optimum.Consequently,the lack of climate-mediated migration at the treeline should raise concerns about the threats posed by warming,such as drought damages,and wildfire,especially in the Mediterranean region.Therefore,conservation management should discuss options and needs to support adaptive management.

Machine learning-based crop recognition from aerial remote sensing imagery

Timely and accurate acquisition of crop distribution and planting area information is important for making agricultural planning and management decisions.This study employed aerial imagery as a data source and machine learning as a classification tool to statically and dynamically identify crops over an agricultural cropping area.Comparative analysis of pixel-based and object-based classifications was performed and classification results were further refined based on three types of object features(layer spectral,geometry,and texture).Static recognition using layer spectral features had the highest accuracy of 75.4%in object-based classification,and dynamic recognition had the highest accuracy of 88.0%in object-based classification based on layer spectral and geometry features.Dynamic identification could not only attenuate the effects of variations on planting dates and plant growth conditions on the results,but also amplify the differences between different features.Object-based classification produced better results than pixel-based classification,and the three feature sets(layer spectral alone,layer spectral and geometry,and all three)resulted in only small differences in accuracy in object-based classification.Dynamic recognition combined with objectbased classification using layer spectral and geometry features could effectively improve crop classification accuracy with high resolution aerial imagery.The methodologies and results from this study should provide practical guidance for crop identification and other agricultural mapping applications.