深度学习网络支持下的农房侵占耕地自动化监测

近年来,违法占用耕地现象屡禁不止,如何利用人工智能等新一代信息技术,快速摸清农村乱占耕地建房底数,做到"早发现、早制止、严查处",是当前整治农村乱占耕地建房工作的研究难点之一。本文通过对高分辨率自然资源影像数据进行预处理,构建基于深度学习网络的自动化监测模型,应用模型进行预测并对输出结果进行GIS优化和空间叠加。试验结果表明,该方法可以快速监测出疑似侵占耕地的违法房屋,为坚守"耕地红线不突破"的底线提供了智能化技术选择,可服务于整治农村乱占耕地建房工作。     

Sentinel-2A数据支持下的雷州半岛植被类型识别

本文以雷州半岛为研究区,利用Sentinel-2A影像数据和真实植被样本数据,综合探讨了机器学习中随机森林与支持向量机的分类效果,并与传统的最大似然法进行比较。提取Sentinel-2A影像9个波段、7个植被指数、72个纹理特征,通过递归特征消除法挑选了10个特征组合,并将其应用于3种分类方法中,对其分类效果进行比较。结果表明:①有效使用多种特征变量是提高植被类型识别精度的关键,就不同特征对植被类型识别的重要性而言,光谱特征与纹理特征相当且大于植被指数,三者重要性相差不大;②随机森林分类效果最佳,不但能对特征进行有效选择,而且能保证植被类型提取精度,提高运行效率;③基于随机森林特征选择的递归特征消除法得到的特征组合不能对其他分类器性能进行优化,对随机森林模型本身的优化效果也有限。     

CatBoost模型在水深反演中的应用

在多光谱遥感水深反演研究中,由于影响反演精度的因素较多,传统的水深反演模型具有一定局限性。机器学习算法在解决非线性高复杂问题上较有优势,将其应用在某些特定区域水深反演可提高反演精度。本文利用Sentinel-2多光谱遥感影像和LiDAR测深数据,以瓦胡岛为研究区域,构建CatBoost水深反演模型,与传统水深反演模型及Boosting中的XGBoost和LightGBM模型的反演精度进行比较。试验结果表明,经过参数优化后的CatBoost水深反演模型的决定系数、均方根误差、平均绝对误差和平均相对误差分别为96.19%、1.09 m、0.77 m和9.61%,准确性最高,效果更佳。     

高光谱和LiDAR联合反演森林生物量研究

估算森林地上生物量(AGB)对于全球实现碳中和目标至关重要。本文以美国缅因州Howland森林为研究区域,借助地面实测样地数据,对比分析协同不同数据源(高光谱和LiDAR)和机器学习算法(随机森林、支持向量机、梯度提升决策树和K最邻近回归)的研究,以改善Howland森林的生物量估计精度。结果表明,采用LiDAR和高光谱植被指数变量模型的最佳精度分别为0.874和0.868,协同高光谱和LiDAR变量并采用梯度提升决策树回归模型的精度为0.927,即多源遥感数据要优于单一数据源。高光谱和LiDAR数据的协同使用对于提高类似于Howland地区或更广泛区域的生物量估计的准确性,具有普遍的适用性与一定的应用前景。     

基于生成对抗网络的建筑物损毁检测EI北大核心CSCD

建筑物作为自然灾害中最受影响的承灾体之一,其损毁信息的准确提取对灾后应急救援具有十分重要的意义。本文借鉴多模态的思想,提出了一种自动检测损毁建筑物的recursive-generative adversarial networks(RS-GAN)方法,将损毁建筑物检测分为灾前建筑物识别和灾后损毁建筑物检测两个任务,且分别在两个GAN分支中完成。RS-GAN加入联合损失函数将两个GAN分支进行连接,充分利用两个任务之间的潜在互利性提升检测效果。RS-GAN利用第1条GAN分支识别建筑物灾前形状与位置,并将识别结果作为第2条GAN分支的输入进行损毁建筑物检测任务,从而使检测结果具有更清晰的轮廓。该方法为端到端模型,在不需要过多的人工干预情形下,实现了损毁建筑物的自动检测。为了验证RS-GAN模型的效果,在圣罗莎和密苏里两个数据集上进行了测试。试验结果表明,RS-GAN方法拥有更好的检测性能,在圣罗莎数据集上的总体精度和平均精度分别达到了0.90和0.86。     

Detail focused fusion of hyperspectral and multispectral imagesEI北大核心CSCD

Hyperspectral image (HSI) and multispectral image (MSI) are two types of images widely used in the field of remote sensing. These images are useful in certain applications, such as environmental monitoring, target detection, and mineral exploration. HSI contains a large amount of spectral information. Photons are typically collected in a larger spatial area on the sensor to ensure a sufficiently high signal-to-noise ratio (SNR). Accordingly, the HSI spatial resolution is much lower compared with MSI. This low spatial resolution greatly affects the practicality of HSI. Accordingly, fusing a low-spatial resolution HSI (LR-HSI) with a high-spatial resolution MSI (HR-MSI) in the same scene to obtain a high-resolution HSI (HR-HSI) is a method for solving such problems, which resolves the contradiction that the spatial resolution and the spectral resolution cannot simultaneously maintain a high level. From the analysis of fusion effect, the spatial and spectral reconstruction errors of the existing algorithms are mainly reflected in the edge and detail areas. The method proposed in this work was a fusion algorithm for dictionary construction and image reconstruction based on detail attention. In terms of maintaining spectral characteristics, the spectral distribution in the detail area is complex and diverse because of the proximity effect of the image. This work proposes to perform dictionary learning on the image and detail layers. The detail perception error terms and a constraint of edge adaptive directional total variation are proposed for spatial characteristic enhancement, which is combined with a local low rank constraint in the same fusion framework to estimate the sparse coefficient. Experiments were conducted on two datasets, namely, Pavia University and Indian Pine, to verify the effectiveness of the proposed method. The quantitative evaluation metrics contain peak SNR, relative dimensionless global error in synthesis, spectral angle map, and universal image quality index. Based on the experimental comparison, the fusion result of the algorithm proposed in this work is significantly improved compared with those of the other algorithms in terms of spatial and spectral characteristics. This work uses dictionary learning to propose a fusion algorithm for dictionary construction and image reconstruction with attention to details through the analysis of the existing hyperspectral and multispectral image fusion algorithms. A hierarchical dictionary learning algorithm is proposed to address the problem of large reconstruction error in the detail part of the existing algorithms. The detail perception error term and the direction adaptive full variational regularization term are used to improve the spectral dictionary solution and coefficient estimation, respectively. The result of the fusion is the error in the spectral characteristics and spatial texture of the detail, which achieves an accurate representation of the edge detail. © 2022 National Remote Sensing Bulletin. All rights reserved.     

Science with society: Evidence-based guidance for best practices in environmental transdisciplinary work

Transdisciplinary research is a promising approach to address sustainability challenges arising from global environmental change, as it is characterized by an iterative process that brings together actors from multiple academic fields and diverse sectors of society to engage in mutual learning with the intent to co-produce new knowledge. We present a conceptual model to guide the implementation of environmental transdisciplinary work, which we consider a “science with society” (SWS) approach, providing suggested activities to conduct throughout a seven-step process. We used a survey with 168 respondents involved in environmental transdisciplinary work worldwide to evaluate the relative importance of these activities and the skills and characteristics required to implement them successfully, with attention to how responses differed according to the gender, geographic location, and positionality of the respondents. Flexibility and collaborative spirit were the most frequently valued skills in SWS, though non-researchers tended to prioritize attributes like humility, trust, and patience over flexibility. We also explored the relative significance of barriers to successful SWS, finding insufficient time and unequal power dynamics were the two most significant barriers to successful SWS. Together with case studies of respondents’ most successful SWS projects, we create a toolbox of 20 best practices that can be used to overcome barriers and increase the societal and scientific impacts of SWS projects. Project success was perceived to be significantly higher where there was medium to high policy impact, and projects initiated by practitioners/other stakeholders had a larger proportion of high policy impact compared to projects initiated by researchers only. Communicating project results to academic audiences occurred more frequently than communicating results to practitioners or the public, despite this being ranked less important overall. We discuss how these results point to three recommendations for future SWS: 1) balancing diverse perspectives through careful partnership formation and design; 2) promoting communication, learning, and reflexivity (i.e., questioning assumptions, beliefs, and practices) to overcome conflict and power asymmetries; and 3) increasing policy impact for joint science and society benefits. Our study highlights the benefits of diversity in SWS - both in the types of people and knowledge included as well as the methods used - and the potential benefits of this approach for addressing the increasingly complex challenges arising from global environmental change.     

Simulation games as a catalyst for social learning: The case of the water-food-energy nexus game

Role-playing simulations have gained in popularity in recent years as a novel method of engaging researchers and stakeholders in a variety of social and environmental issues. While academic interest has grown on this topic, knowledge remains sparse on the underlying theories that may guide the design of such games. Thsi article introduces a new game design framework - Com­pleC­Sus (Com­plex­ity-Col­lab­o­ra­tion-Sus­tain­abil­ity) - built on the concepts of social learning and procedural rhetoric. We describe and discuss the conceptual basis for our framework, giving a detailed account of its application through the recently developed the Water–Food–Energy Nexus Game (Nexus Game) as an example. We illustrate the process involved in designing the Nexus Game through initial scoping, prototyping, and design decisions, and how game structure and debriefing have been crafted to foster social learning focused on the understanding of the underlying social-ecological system as well as fostering collaboration between stakeholders. We also provide the analysis of qualitative data collected during recent gaming sessions across three continents to evaluate the Nexus Game’s potential learning effects.     

基于数据驱动高阶学习律的轮式移动机器人轨迹跟踪控制

轮式机器人执行巡逻、播种和工业生产等任务是一个强非线性的间歇过程.针对重复运行的轮式机器人轨迹跟踪问题,本文提出了一种基于数据驱动的高阶迭代学习控制算法.首先,对轮式移动机器人的模型进行推导设计,并对推导得到的状态空间形式的离散时间模型利用基于状态转移的迭代动态线性化方法,将轮式机器人系统转化为线性输入输出数据模型;其次,设计高阶迭代优化目标函数得到控制律,并利用参数更新律估计线性输入输出数据模型中的未知参数.控制器的设计和分析只使用系统的输入输出数据,不包含任何显式的模型信息.通过采用高阶学习控制方法,在控制律中利用更多之前迭代的控制输入信息,提高了控制性能.最后,仿真结果验证了该方法在轮式机器人轨迹跟踪控制中的有效性.     

Snow avalanche susceptibility evaluation in the central Shaluli Mountains of Tibetan Plateau based on machine learning method北大核心CSCD

Snow avalanches,which are widely and frequently developed at high elevations,seriously threatens the built traffic corridors in the Tibetan Plateau. Susceptibility evaluation of snow avalanche via machine learning model with a high forecast accuracy can be appled to quickly and effectively assess the regional avalanche risk. This paper took the central Shaluli Mountain region as the study area,in which the snow avalanche inventory was established through remote sensing interpretation and field investigation verification. We quantitatively extracted 17 evaluation factors via GIS-based analysis,and these factors were selected through the variance expansion factor（VIF）. Four machine learning models containing SVM,DT,MLP and KNN were used to compile the susceptibility index map of snow avalanches,and kappa coefficient and ROC curve were used to verify the accuracy. The results suggested that the susceptibility indexes obtained from SVM,DT,MLP and KNN were in the range of［0,0. 964］,［0,815］,［0,0. 995］and［0,1］,respectively. The accuracy test results show that these four models all have good prediction accuracy. Among them,the SVM model is the best. The results also indicated that the areas with the high snow avalanche susceptibility mainly distributed in Genie Mountain and Rigong Mountain,most of which were above the planation surface of the Tibetan Plateau. The average altitude of the extremely high snow-avalanche-prone areas is 4 939 m,while the average altitude of the high snow avalanche-prone areas is 4 859 m. The snow avalanche has low perniciousness on the Sichuan-Tibet Highway and the Sichuan-Tibet Railway in the study area. This study can provide theoretical basis and method reference for disaster prevention and mitigation of snow avalanche along Sichuan-Tibet Railway and other major projects across Shaluli Mountains region. © 2022 Science Press (China).