Snow cover plays a critical role in global climate regulation and hydrological processes.Accurate monitoring is essential for understanding snow distribution patterns,managing water resources,and assessing the impacts...Snow cover plays a critical role in global climate regulation and hydrological processes.Accurate monitoring is essential for understanding snow distribution patterns,managing water resources,and assessing the impacts of climate change.Remote sensing has become a vital tool for snow monitoring,with the widely used Moderate-resolution Imaging Spectroradiometer(MODIS)snow products from the Terra and Aqua satellites.However,cloud cover often interferes with snow detection,making cloud removal techniques crucial for reliable snow product generation.This study evaluated the accuracy of four MODIS snow cover datasets generated through different cloud removal algorithms.Using real-time field camera observations from four stations in the Tianshan Mountains,China,this study assessed the performance of these datasets during three distinct snow periods:the snow accumulation period(September-November),snowmelt period(March-June),and stable snow period(December-February in the following year).The findings showed that cloud-free snow products generated using the Hidden Markov Random Field(HMRF)algorithm consistently outperformed the others,particularly under cloud cover,while cloud-free snow products using near-day synthesis and the spatiotemporal adaptive fusion method with error correction(STAR)demonstrated varying performance depending on terrain complexity and cloud conditions.This study highlighted the importance of considering terrain features,land cover types,and snow dynamics when selecting cloud removal methods,particularly in areas with rapid snow accumulation and melting.The results suggested that future research should focus on improving cloud removal algorithms through the integration of machine learning,multi-source data fusion,and advanced remote sensing technologies.By expanding validation efforts and refining cloud removal strategies,more accurate and reliable snow products can be developed,contributing to enhanced snow monitoring and better management of water resources in alpine and arid areas.展开更多
This paper develops a variational model for image noise removal using total curvature(TC), which is a high-order regularizer. The TC has the advantage of preserving image feature. Unfortunately, it also has the charac...This paper develops a variational model for image noise removal using total curvature(TC), which is a high-order regularizer. The TC has the advantage of preserving image feature. Unfortunately, it also has the characteristics of nonlinear, non-convex and non-smooth. Consequently, the numerical computation with the curvature regularization is difficult. In order to conquer the computation problem, the proposed model is transformed into an alternating optimization problem by importing auxiliary variables. Furthermore, based on alternating direction method of multipliers, we design a fast numerical approximation iterative scheme for proposed model. Finally, numerous experiments are implemented to indicate the advantages of the proposed model in image edge preserving, image contrast and corners preserving. Meanwhile, the high computational efficiency of the designed model is verified by comparing with traditional models, including the total variation(TV) and total Laplace(TL) model.展开更多
Aiming at low SNR and feature distortion of valid neural signals caused by eye movement,electromagnetic interference and channel noise in frontal single/multi-channel BCI EEG tests,a BIAS-core anchor artifact removal ...Aiming at low SNR and feature distortion of valid neural signals caused by eye movement,electromagnetic interference and channel noise in frontal single/multi-channel BCI EEG tests,a BIAS-core anchor artifact removal algorithm is proposed for low-channel,low-computing wearable scenarios.It defines EOG-BIAS,ENV-BIAS,CH-BIAS,builds a"pre-calibration-real-time processing-posterior compensation"lightweight architecture,and decouples artifacts via"bias modeling-separation-compensation".With unsupervised learning and individual adaptability,it fits unshielded real-time environments.Tests show its EOG/ENV-BIAS removal rates≥95%/≥99%,neural signal retention≥90%,single-frame delay<10ms,SNR improvement≥15dB(stable≥10dB),effectively ensuring frontal BCI signal quality and supporting low-resource frontal BCI signal preprocessing.展开更多
Recent studies of the space debris environment in Low Earth Orbit(LEO)have shown that the critical density of space debris has been reached in certain regions.The Active Debris Removal(ADR)mission,to mitigate the spac...Recent studies of the space debris environment in Low Earth Orbit(LEO)have shown that the critical density of space debris has been reached in certain regions.The Active Debris Removal(ADR)mission,to mitigate the space debris density and stabilize the space debris environment,has been considered as a most effective method.In this paper,a novel two-level optimization strategy for multi-debris removal mission in LEO is proposed,which includes the low-level and high-level optimization process.To improve the overall performance of the multi-debris active removal mission and obtain multiple Pareto-optimal solutions,the ADR mission is seen as a Time-Dependant Traveling Salesman Problem(TDTSP)with two objective functions to minimize the total mission duration and the total propellant consumption.The problem includes the sequence optimization to determine the sequence of removal of space debris and the transferring optimization to define the orbital maneuvers.Two optimization models for the two-level optimization strategy are built in solving the multi-debris removal mission,and the optimal Pareto solution is successfully obtained by using the non-dominated sorting genetic algorithm II(NSGA-II).Two test cases are presented,which show that the low level optimization strategy can successfully obtain the optimal sequences and the initial solution of the ADR mission and the high level optimization strategy can efficiently and robustly find the feasible optimal solution for long duration perturbed rendezvous problem.展开更多
In this paper,a new mission model,called a multi-debris active removal mission with partial debris capture strategy,is proposed.The model assumes that a platform only captures part of the scheduled debris at a time an...In this paper,a new mission model,called a multi-debris active removal mission with partial debris capture strategy,is proposed.The model assumes that a platform only captures part of the scheduled debris at a time and then releases these debris pieces to a disposal orbit.This process is then repeated until all of the scheduled debris is removed.A genetic algorithm with a multiparameter concatenated coding method is designed to optimize the plan of a multi-debris active removal mission with a partial debris capture strategy.A set of six pieces of debris and a set of 10 pieces of debris are selected to demonstrate the proposed planning method.The result confirms the effectiveness of the genetic algorithm with the multi-parameter concatenated coding method.The new mission model provides a more comprehensive decision-making framework than the existing mission models and makes it possible to further decrease mission costs.展开更多
为解决室内视觉定位中因家具、行人等遮挡物体导致的定位精度下降问题,提出了一种基于遮挡清除算法的室内视觉定位系统。系统首先利用YOLOv11目标检测算法识别并定位遮挡物体,对遮挡区域像素进行重置并生成掩码图像;随后采用MAT(mask-aw...为解决室内视觉定位中因家具、行人等遮挡物体导致的定位精度下降问题,提出了一种基于遮挡清除算法的室内视觉定位系统。系统首先利用YOLOv11目标检测算法识别并定位遮挡物体,对遮挡区域像素进行重置并生成掩码图像;随后采用MAT(mask-aware transformer)图像修复算法恢复无遮挡图像,以获取完整的环境信息用于后续定位。在定位阶段,基于本质矩阵的定位算法结合SuperPoint特征提取与LightGlue特征匹配方法,有效提升了在光照变化和视点偏移下的鲁棒性。实验结果表明,提出的系统在7Scenes数据集和自制LH数据集上表现优异。在7Scenes上,定位误差为(0.063 m, 1.759°),相比室内视觉定位领域最广泛应用于工业上SuperPoint+SuperGlue算法,位置精度提升约6%,旋转误差降低约4%;在LH数据集上,位置精度进一步提升约15%。在与最先进的算法对比时,尽管精度略逊一筹,但考虑到该方法使用的是去遮挡后的图像数据集,其在复杂场景中的表现依然出色,尤其在处理遮挡问题时能够保持较高的鲁棒性和精度。此外,该方法更具轻量性,能够更快速地部署和训练,适用于实时应用场景。展开更多
Some depression cells with heights lower than their surrounding cells may often be found in Grid-based digital elevation models (DEM) dataset due to sampling errors.The depression-filling algorithm presented by Planch...Some depression cells with heights lower than their surrounding cells may often be found in Grid-based digital elevation models (DEM) dataset due to sampling errors.The depression-filling algorithm presented by Planchon and Darboux works very quickly compared to other published methods.Despite its simplicity and deli-cacy,this algorithm remains difficult to understand due to its three complex subroutines and its recursive execution.Another fast algorithm is presented in this article.The main idea of this new algorithm is as follows:first,the DEM dataset is viewed as an island and the outer space as an ocean;when the ocean level increases,the DEM cells on the island's boundary will be inundated;when a cell is inundated for the first time,its elevation is increased to the ocean level at that moment;after the ocean has inun-dated the entire DEM,all of the depressions are filled.The depression-removing processing is performed using a priority queue.Theoretically,this new algorithm is a fast algorithm despite the fact that it runs more slowly than Planchon and Darboux's method.Its time-complexity in both the worst case and in an average case is O(8nlog 2 (m)),which is close to O(n).The running speed of this algorithm depends mainly on the insertion operation of the priority queue.As shown by the tests,the depres-sion-filling effects of this algorithm are correct and valid,and the overall time consumption of this algorithm is less than twice the time consumed by Planchon & Darboux's method for handling a DEM smaller than 2500×2500 cells.More importantly,this new algorithm is simpler and easier to understand than Planchon and Darboux's method This advantage allows the correct program code to be written quickly.展开更多
云污染干扰了遥感卫星对地面信息的获取,去云算法的开发有效解决了由厚云遮挡所引起的遥感数据缺失问题。然而,现有算法通过计算中心像元与邻近时空像元之间的差异,定义相似像元,在地表突变的情况下,这种选取方法面临逻辑问题。因此提...云污染干扰了遥感卫星对地面信息的获取,去云算法的开发有效解决了由厚云遮挡所引起的遥感数据缺失问题。然而,现有算法通过计算中心像元与邻近时空像元之间的差异,定义相似像元,在地表突变的情况下,这种选取方法面临逻辑问题。因此提出一种顾及时间变化的相似像元选取方法以增强现有去云算法。利用K-means聚类提取时序清洁像元的时间变化聚类中心,进而利用这些聚类中心对清洁时期影像进行监督分类,推算云下时序变化模式,认为与中心像元变化模式相同的像元具有相同的时序变化,进而与现有算法寻找的相似像元求交集得到优选的相似像元,按照原有去云算法完成去云。选择豫北农区作为试验区,以改进的相似像元插值法(modified neighborhood similar pixel interpolator,MNSPI)和加权线性回归(weighted linear regression,WLR)2个去云算法为例,结果表明:①优选算法可以从快速变化的农事情景中准确提取相似像元;②基于优选结果的去云算法在光谱差异以及结构相似度都优于原算法,改进之后的MNSPI算法平均均方根误差(root mean squared error,RMSE)从0.0234下降到0.0152,结构相似性指数(structural similarity index measure,SSIM)从0.6023提升为0.7166;改进后的WLR算法平均RMSE从0.0375下降到0.0168,SSIM从0.5897提升到0.6461;③实际影像的应用表明算法可以准确恢复云下信息,视觉效果良好。研究可以为快速变化地表的遥感图像去云提供科学依据,为高质量遥感时序监测提供数据支撑。展开更多
基金funded by the Third Xinjiang Scientific Expedition Program(2021xjkk1400)the National Natural Science Foundation of China(42071049)+2 种基金the Natural Science Foundation of Xinjiang Uygur Autonomous Region(2019D01C022)the Xinjiang Uygur Autonomous Region Innovation Environment Construction Special Project&Science and Technology Innovation Base Construction Project(PT2107)the Tianshan Talent-Science and Technology Innovation Team(2022TSYCTD0006).
摘要Snow cover plays a critical role in global climate regulation and hydrological processes.Accurate monitoring is essential for understanding snow distribution patterns,managing water resources,and assessing the impacts of climate change.Remote sensing has become a vital tool for snow monitoring,with the widely used Moderate-resolution Imaging Spectroradiometer(MODIS)snow products from the Terra and Aqua satellites.However,cloud cover often interferes with snow detection,making cloud removal techniques crucial for reliable snow product generation.This study evaluated the accuracy of four MODIS snow cover datasets generated through different cloud removal algorithms.Using real-time field camera observations from four stations in the Tianshan Mountains,China,this study assessed the performance of these datasets during three distinct snow periods:the snow accumulation period(September-November),snowmelt period(March-June),and stable snow period(December-February in the following year).The findings showed that cloud-free snow products generated using the Hidden Markov Random Field(HMRF)algorithm consistently outperformed the others,particularly under cloud cover,while cloud-free snow products using near-day synthesis and the spatiotemporal adaptive fusion method with error correction(STAR)demonstrated varying performance depending on terrain complexity and cloud conditions.This study highlighted the importance of considering terrain features,land cover types,and snow dynamics when selecting cloud removal methods,particularly in areas with rapid snow accumulation and melting.The results suggested that future research should focus on improving cloud removal algorithms through the integration of machine learning,multi-source data fusion,and advanced remote sensing technologies.By expanding validation efforts and refining cloud removal strategies,more accurate and reliable snow products can be developed,contributing to enhanced snow monitoring and better management of water resources in alpine and arid areas.
基金supported by the National Natural Science Foundation of China(No.61602269)the China Postdoctoral Science Foundation(No.2015M571993)+1 种基金the Shandong Provincial Natural Science Foundation of China(No.ZR2017MD004)the Qingdao Postdoctoral Application Research Funded Project
摘要This paper develops a variational model for image noise removal using total curvature(TC), which is a high-order regularizer. The TC has the advantage of preserving image feature. Unfortunately, it also has the characteristics of nonlinear, non-convex and non-smooth. Consequently, the numerical computation with the curvature regularization is difficult. In order to conquer the computation problem, the proposed model is transformed into an alternating optimization problem by importing auxiliary variables. Furthermore, based on alternating direction method of multipliers, we design a fast numerical approximation iterative scheme for proposed model. Finally, numerous experiments are implemented to indicate the advantages of the proposed model in image edge preserving, image contrast and corners preserving. Meanwhile, the high computational efficiency of the designed model is verified by comparing with traditional models, including the total variation(TV) and total Laplace(TL) model.
基金the Open Research Foundation of Science and Technology in Aerospace Flight Dynamics Laboratory of China(GF2018005).
摘要Recent studies of the space debris environment in Low Earth Orbit(LEO)have shown that the critical density of space debris has been reached in certain regions.The Active Debris Removal(ADR)mission,to mitigate the space debris density and stabilize the space debris environment,has been considered as a most effective method.In this paper,a novel two-level optimization strategy for multi-debris removal mission in LEO is proposed,which includes the low-level and high-level optimization process.To improve the overall performance of the multi-debris active removal mission and obtain multiple Pareto-optimal solutions,the ADR mission is seen as a Time-Dependant Traveling Salesman Problem(TDTSP)with two objective functions to minimize the total mission duration and the total propellant consumption.The problem includes the sequence optimization to determine the sequence of removal of space debris and the transferring optimization to define the orbital maneuvers.Two optimization models for the two-level optimization strategy are built in solving the multi-debris removal mission,and the optimal Pareto solution is successfully obtained by using the non-dominated sorting genetic algorithm II(NSGA-II).Two test cases are presented,which show that the low level optimization strategy can successfully obtain the optimal sequences and the initial solution of the ADR mission and the high level optimization strategy can efficiently and robustly find the feasible optimal solution for long duration perturbed rendezvous problem.
基金co-supported by the Open Fund Project of Space Intelligent Control Technology Laboratory(No.HTKJ2021KL502010)the Research Project of Space Debris and Near-earth Asteroid Defense Grants,China(No.KJSP 2020010303)the National Natural Science Foundation of China(No.11802130).
摘要In this paper,a new mission model,called a multi-debris active removal mission with partial debris capture strategy,is proposed.The model assumes that a platform only captures part of the scheduled debris at a time and then releases these debris pieces to a disposal orbit.This process is then repeated until all of the scheduled debris is removed.A genetic algorithm with a multiparameter concatenated coding method is designed to optimize the plan of a multi-debris active removal mission with a partial debris capture strategy.A set of six pieces of debris and a set of 10 pieces of debris are selected to demonstrate the proposed planning method.The result confirms the effectiveness of the genetic algorithm with the multi-parameter concatenated coding method.The new mission model provides a more comprehensive decision-making framework than the existing mission models and makes it possible to further decrease mission costs.
摘要为解决室内视觉定位中因家具、行人等遮挡物体导致的定位精度下降问题,提出了一种基于遮挡清除算法的室内视觉定位系统。系统首先利用YOLOv11目标检测算法识别并定位遮挡物体,对遮挡区域像素进行重置并生成掩码图像;随后采用MAT(mask-aware transformer)图像修复算法恢复无遮挡图像,以获取完整的环境信息用于后续定位。在定位阶段,基于本质矩阵的定位算法结合SuperPoint特征提取与LightGlue特征匹配方法,有效提升了在光照变化和视点偏移下的鲁棒性。实验结果表明,提出的系统在7Scenes数据集和自制LH数据集上表现优异。在7Scenes上,定位误差为(0.063 m, 1.759°),相比室内视觉定位领域最广泛应用于工业上SuperPoint+SuperGlue算法,位置精度提升约6%,旋转误差降低约4%;在LH数据集上,位置精度进一步提升约15%。在与最先进的算法对比时,尽管精度略逊一筹,但考虑到该方法使用的是去遮挡后的图像数据集,其在复杂场景中的表现依然出色,尤其在处理遮挡问题时能够保持较高的鲁棒性和精度。此外,该方法更具轻量性,能够更快速地部署和训练,适用于实时应用场景。
基金financially supported by the National Basic Research Program of China (Grant No.2006CB400502)the Promotion of 100 Young Talent Scientist Project of the Chinese Acad-emy of Sciences (8-057493)the Special Meteorology Project(GYHY(QX)2007-6-1)
摘要Some depression cells with heights lower than their surrounding cells may often be found in Grid-based digital elevation models (DEM) dataset due to sampling errors.The depression-filling algorithm presented by Planchon and Darboux works very quickly compared to other published methods.Despite its simplicity and deli-cacy,this algorithm remains difficult to understand due to its three complex subroutines and its recursive execution.Another fast algorithm is presented in this article.The main idea of this new algorithm is as follows:first,the DEM dataset is viewed as an island and the outer space as an ocean;when the ocean level increases,the DEM cells on the island's boundary will be inundated;when a cell is inundated for the first time,its elevation is increased to the ocean level at that moment;after the ocean has inun-dated the entire DEM,all of the depressions are filled.The depression-removing processing is performed using a priority queue.Theoretically,this new algorithm is a fast algorithm despite the fact that it runs more slowly than Planchon and Darboux's method.Its time-complexity in both the worst case and in an average case is O(8nlog 2 (m)),which is close to O(n).The running speed of this algorithm depends mainly on the insertion operation of the priority queue.As shown by the tests,the depres-sion-filling effects of this algorithm are correct and valid,and the overall time consumption of this algorithm is less than twice the time consumed by Planchon & Darboux's method for handling a DEM smaller than 2500×2500 cells.More importantly,this new algorithm is simpler and easier to understand than Planchon and Darboux's method This advantage allows the correct program code to be written quickly.
摘要云污染干扰了遥感卫星对地面信息的获取,去云算法的开发有效解决了由厚云遮挡所引起的遥感数据缺失问题。然而,现有算法通过计算中心像元与邻近时空像元之间的差异,定义相似像元,在地表突变的情况下,这种选取方法面临逻辑问题。因此提出一种顾及时间变化的相似像元选取方法以增强现有去云算法。利用K-means聚类提取时序清洁像元的时间变化聚类中心,进而利用这些聚类中心对清洁时期影像进行监督分类,推算云下时序变化模式,认为与中心像元变化模式相同的像元具有相同的时序变化,进而与现有算法寻找的相似像元求交集得到优选的相似像元,按照原有去云算法完成去云。选择豫北农区作为试验区,以改进的相似像元插值法(modified neighborhood similar pixel interpolator,MNSPI)和加权线性回归(weighted linear regression,WLR)2个去云算法为例,结果表明:①优选算法可以从快速变化的农事情景中准确提取相似像元;②基于优选结果的去云算法在光谱差异以及结构相似度都优于原算法,改进之后的MNSPI算法平均均方根误差(root mean squared error,RMSE)从0.0234下降到0.0152,结构相似性指数(structural similarity index measure,SSIM)从0.6023提升为0.7166;改进后的WLR算法平均RMSE从0.0375下降到0.0168,SSIM从0.5897提升到0.6461;③实际影像的应用表明算法可以准确恢复云下信息,视觉效果良好。研究可以为快速变化地表的遥感图像去云提供科学依据,为高质量遥感时序监测提供数据支撑。