面向遥感图像场景分类的LAG-MANet模型

遥感图像分类过程中,局部信息与全局信息至关重要。目前,遥感图像分类的方法主要包括卷积神经网络(CNN)及Transformer。CNN在局部信息提取方面具有优势,但在全局信息提取方面有一定的局限性。相比之下,Transformer在全局信息提取方面表现出色,但计算复杂度高。为提高遥感图像场景分类性能,降低复杂度,设计了LAG-MANet纯卷积网络。该网络既关注局部特征,又关注全局特征,并且考虑了多尺度特征。输入图像被预处理后,首先采用多分支扩张卷积模块(MBDConv)提取多尺度特征;然后依次进入网络的4个阶段,在每个阶段采用并行双域特征融合模块(P2DF)分支路提取局部、全局特征并进行融合;最后先经过全局平均池化、再经过全连接层输出分类标签。LAG-MANet在WHU-RS19数据集、SIRI-WHU数据集及RSSCN7数据集上的分类准确率分别为97.76%、97.04%、97.18%。试验结果表明,在3个具有挑战性的公开遥感数据集上,LAG-MANet更具有优越性。

基于改进HRNet的遥感影像冬小麦语义分割方法

冬小麦在影像中呈现田块碎小且分布零散等空间特征,同时影像包含的复杂地物对冬小麦识别造成干扰,易出现识别精度低且边界分割模糊等问题。为及时准确获取大范围冬小麦空间分布信息,该研究以高分二号卫星影像作为数据源,提出一种CAHRNet(change attention high-resolution Net)语义分割模型。采用HRNet(high-resolution Net)替换ResNet作为模型的主干网络,网络的并行交互方式易获取高分辨率的特征信息;联合OCR(object-contextual representations)模块聚合上下文信息,以增强像素点与目标对象区域的关联性;3)引入坐标注意力(coordinate attention)机制,使网络模型充分利用有效的空间位置信息,以保留分割区域的边缘细节,提高对分布零散、形状多变的冬小麦田块的特征提取能力。试验结果表明,在自制的高分辨率遥感数据集上,CAHRNet模型的平均交并比(mean intersection over union,MIoU)和像素准确率(pixel accuracy, PA)分别达到81.72%和97.08%,MIoU相较U-Net、DeepLabv3+分别提高了9.09、2.44个百分点;PA相较U-Net、DeepLabv3+分别提高6.80、1.59个百分点,说明CAHRNet模型具有较高的分割识别精度,可为进一步准确获取冬小麦作物分布信息提供技术支撑。

基于OneNET云平台的智慧消防远程监控系统的设计

针对传统消防监控系统存在开发成本高、误警率高、实时监控不便的问题,提出一种基于物联网云平台的智慧消防远程监控系统。采用STM32单片机作为中枢控制芯片,经多传感器采集温度、湿度、烟雾、火焰等环境数据,通过窄带物联网(NB-IoT,Narrow Band Internet of Things)上传至OneNET云平台。经数据分析后以可视化方式呈现,对异常数据触发报警实时响应。通过手机APP实现数据实时监测及一键处置。经测试,监控系统报警准确率高于97.2%,数据延迟低于50 ms,表明该系统能够实现消防火警的无线远程监控,并做出快速反应,满足中小微企业和普通家庭用户的消防监控需要。

基于HRNet和自注意力机制的多源遥感影像水稻提取

为了提高协同光学和雷达影像提取水稻的精度,该研究通过改进深度学习网络HRNet,提出一种多级特征融合的框架方法,改进后的MSATT-HRNet模型综合利用欧空局哥白尼项目哨兵1号(Sentinel-1)的双极化合成孔径雷达(synthetic aperture radar,SAR)影像和哨兵2号(Sentinel-2)多光谱光学影像,提取了湖南省长沙市望城区水稻种植区域。针对HRNet网络的改进主要包括两部分:1)设计了通道注意力机制与最大池化组成的卷积组用于提取SAR特征,同时将自注意力模块嵌入HRNet基础特征提取模块中用于提取多光谱光学影像的特征;2)为了探索双模态特征之间的内在互补关系,设计了由通道注意力与空间注意力组成的特征融合模块。研究针对改进模型进行了消融试验,并将MSATT-HRNet与其他常用深度学习方法(MCANet、Deeplabv3、Unet)进行了比较。结果表明,该研究提出的多源数据融合方法能够利用不同数据源的互补优势。水稻种植区域提取结果的总体精度、Kappa系数分别达到97.04%和0.961,与MCANet、Deeplabv3、Unet相比,总体精度分别提高6.90、2.67和2.98个百分点,Kappa系数分别提高0.055、0.025和0.030。证实了该方法可以有效提高水稻的判别精度。研究通过深度学习技术与遥感影像的耦合,为南方多云雨地区水稻制图提供了一种可行的选择。

基于改进U-Net模型的遥感影像道路提取方法研究

针对高分辨率遥感影像背景信息复杂,道路提取难度大,自动化程度低等问题,论文提出了一种改进的U-Net的道路提取方法。首先,编码器使用VGG16网络结构替代原始U-Net编码器结构;然后,在每个编码器和解码器块后加入特征压缩激活模块(SENet)增强网络特征学习能力;最后,使用Dice损失函数和二分类交叉熵损失函数复合的损失函数进行训练,减轻了道路提取任务中的样本不平衡问题。在Massachusetts Road数据集上的结果表明,改进后的算法对道路提取结果得到了有效的提升。所提方法在测试集上的精确度、召回率、F1-score和mIoU评价指标分别达到82.5%、77.8%、80.0%及82.1%,在测试影像中对错综交叉的道路具有更好的识别效果。

基于级联U-Net的遥感影像道路分割和轮廓提取方法

针对基于深度学习的遥感图像道路信息提取模型往往只能输出单任务结果且多任务之间相关性利用不充分的问题,提出了一种基于级联U-Net的道路语义分割和轮廓联合检测方法,将道路语义分割后的特征图与原始图像融合后进行道路轮廓的提取,实现道路语义分割和边界轮廓的联合训练。首先使用U-Net网络结构提取光学遥感图像丰富的层次化特征,通过级联结构将特征串联融合,分别用于提取道路的语义类别和边界轮廓。其次在每级U-Net结构中引入注意力机制模块,进行空间上下文信息和深层次特征提取,改善网络提取过程中出现的细节模糊现象。最后,使用骰子系数和交叉熵误差组成的联合损失函数进行多任务整体训练,实现深度学习模型对遥感图像中道路语义类别和边界轮廓的同时提取。通过在加拿大渥太华城市地区的光学遥感数据集上进行实验,基于级联U-Net的道路信息联合提取方法在分割指标上分别获得了42%的精确度、58%的召回率、48.2%的F1分数以及71.6%的平均交并比,在道路检测指标上取得了0.896的全局最佳阈值(ODS)。结果表明,该模型在满足联合提取道路多任务信息的同时具有更优的检测精度。

ConvNeXt-UperNet-Based Deep Learning Model for Road Extraction from High-Resolution Remote Sensing Images

When existing deep learning models are used for road extraction tasks from high-resolution images,they are easily affected by noise factors such as tree and building occlusion and complex backgrounds,resulting in incomplete road extraction and low accuracy.We propose the introduction of spatial and channel attention modules to the convolutional neural network ConvNeXt.Then,ConvNeXt is used as the backbone network,which cooperates with the perceptual analysis network UPerNet,retains the detection head of the semantic segmentation,and builds a new model ConvNeXt-UPerNet to suppress noise interference.Training on the open-source DeepGlobe and CHN6-CUG datasets and introducing the DiceLoss on the basis of CrossEntropyLoss solves the problem of positive and negative sample imbalance.Experimental results show that the new network model can achieve the following performance on the DeepGlobe dataset:79.40%for precision(Pre),97.93% for accuracy(Acc),69.28% for intersection over union(IoU),and 83.56% for mean intersection over union(MIoU).On the CHN6-CUG dataset,the model achieves the respective values of 78.17%for Pre,97.63%for Acc,65.4% for IoU,and 81.46% for MIoU.Compared with other network models,the fused ConvNeXt-UPerNet model can extract road information better when faced with the influence of noise contained in high-resolution remote sensing images.It also achieves multiscale image feature information with unified perception,ultimately improving the generalization ability of deep learning technology in extracting complex roads from high-resolution remote sensing images.

Advancements in Remote Sensing Image Dehazing: Introducing URA-Net with Multi-Scale Dense Feature Fusion Clusters and Gated Jump Connection

The degradation of optical remote sensing images due to atmospheric haze poses a significant obstacle,profoundly impeding their effective utilization across various domains.Dehazing methodologies have emerged as pivotal components of image preprocessing,fostering an improvement in the quality of remote sensing imagery.This enhancement renders remote sensing data more indispensable,thereby enhancing the accuracy of target iden-tification.Conventional defogging techniques based on simplistic atmospheric degradation models have proven inadequate for mitigating non-uniform haze within remotely sensed images.In response to this challenge,a novel UNet Residual Attention Network(URA-Net)is proposed.This paradigmatic approach materializes as an end-to-end convolutional neural network distinguished by its utilization of multi-scale dense feature fusion clusters and gated jump connections.The essence of our methodology lies in local feature fusion within dense residual clusters,enabling the extraction of pertinent features from both preceding and current local data,depending on contextual demands.The intelligently orchestrated gated structures facilitate the propagation of these features to the decoder,resulting in superior outcomes in haze removal.Empirical validation through a plethora of experiments substantiates the efficacy of URA-Net,demonstrating its superior performance compared to existing methods when applied to established datasets for remote sensing image defogging.On the RICE-1 dataset,URA-Net achieves a Peak Signal-to-Noise Ratio(PSNR)of 29.07 dB,surpassing the Dark Channel Prior(DCP)by 11.17 dB,the All-in-One Network for Dehazing(AOD)by 7.82 dB,the Optimal Transmission Map and Adaptive Atmospheric Light For Dehazing(OTM-AAL)by 5.37 dB,the Unsupervised Single Image Dehazing(USID)by 8.0 dB,and the Superpixel-based Remote Sensing Image Dehazing(SRD)by 8.5 dB.Particularly noteworthy,on the SateHaze1k dataset,URA-Net attains preeminence in overall performance,yielding defogged images characterized by consistent visual quality.This underscores the contribution of the research to the advancement of remote sensing technology,providing a robust and efficient solution for alleviating the adverse effects of haze on image quality.

基于改进STANet的遥感图像变化检测算法

遥感图像变化检测是为了识别出双时相图像之间的显著变化。给定2个在不同时间拍摄的配准图像,光照变化和错配误差会掩盖真实物体的变化,探索不同时空像素之间的关系可以提高遥感图像变化检测方法的性能。在Spatial Temporal Attention Neural Network(STANet)中,提出了一种基于孪生的时空注意力神经网络。在其基础上进行改进:①对距离度量模块由于线性插值导致的变化特征间隙模糊问题,设计了对距离特征的上采样模块,使得变化区域间隙更加明显,虚警率更低;②针对STANet的Pyramid Spatial Temporal Attention Module(PAM)模块计算开销大的问题,引用了新的Coordinate Attention(CA)模块,在降低运算开销的基础上,更好地识别了不同空间、通道的特征;③针对STANet对Residual Network(ResNet)提取出的特征图利用不充分的问题,加入了深监督模块,利用中间层的特征计算一个权重衰减的loss,起到正则化的作用。实验表明,改进之后的网络将基线模型的F1得分从81.6提高到86.1。在公共遥感图像数据集上的实验结果表明,改进的方法优于其他几种先进的方法。

Optimized Binary Neural Networks for Road Anomaly Detection:A TinyML Approach on Edge Devices

Integrating Tiny Machine Learning(TinyML)with edge computing in remotely sensed images enhances the capabilities of road anomaly detection on a broader level.Constrained devices efficiently implement a Binary Neural Network(BNN)for road feature extraction,utilizing quantization and compression through a pruning strategy.The modifications resulted in a 28-fold decrease in memory usage and a 25%enhancement in inference speed while only experiencing a 2.5%decrease in accuracy.It showcases its superiority over conventional detection algorithms in different road image scenarios.Although constrained by computer resources and training datasets,our results indicate opportunities for future research,demonstrating that quantization and focused optimization can significantly improve machine learning models’accuracy and operational efficiency.ARM Cortex-M0 gives practical feasibility and substantial benefits while deploying our optimized BNN model on this low-power device:Advanced machine learning in edge computing.The analysis work delves into the educational significance of TinyML and its essential function in analyzing road networks using remote sensing,suggesting ways to improve smart city frameworks in road network assessment,traffic management,and autonomous vehicle navigation systems by emphasizing the importance of new technologies for maintaining and safeguarding road networks.

Weakly Supervised Network with Scribble-Supervised and Edge-Mask for Road Extraction from High-Resolution Remote Sensing Images

Significant advancements have been achieved in road surface extraction based on high-resolution remote sensingimage processing. Most current methods rely on fully supervised learning, which necessitates enormous humaneffort to label the image. Within this field, other research endeavors utilize weakly supervised methods. Theseapproaches aim to reduce the expenses associated with annotation by leveraging sparsely annotated data, such asscribbles. This paper presents a novel technique called a weakly supervised network using scribble-supervised andedge-mask (WSSE-net). This network is a three-branch network architecture, whereby each branch is equippedwith a distinct decoder module dedicated to road extraction tasks. One of the branches is dedicated to generatingedge masks using edge detection algorithms and optimizing road edge details. The other two branches supervise themodel’s training by employing scribble labels and spreading scribble information throughout the image. To addressthe historical flaw that created pseudo-labels that are not updated with network training, we use mixup to blendprediction results dynamically and continually update new pseudo-labels to steer network training. Our solutiondemonstrates efficient operation by simultaneously considering both edge-mask aid and dynamic pseudo-labelsupport. The studies are conducted on three separate road datasets, which consist primarily of high-resolutionremote-sensing satellite photos and drone images. The experimental findings suggest that our methodologyperforms better than advanced scribble-supervised approaches and specific traditional fully supervised methods.

CNS-Net:一种循环多注意力特征聚合架构

基于卷积神经网络(convolutional neural network,CNN)方法在高分辨率遥感图像分类时,提取全局特征信息和局部特征信息方面有一定优势,但不能有效区别关键信息和干扰信息。由此,提出一种端到端的CNS-Net网络来提取图像显著性特征。首先,设计了一个全局增强性模块(global enhancement module,GEM)通过显示建模通道之间的相互依赖关系,使网络选择性提取关键区域;其次,提出了一个多级循环注意模块(multi-stage cyclic attention module,MCAM)来捕捉特征信息的长期依赖关系和上下文感知信息;最后,在4个公共数据集上进行实验,结果表明所提方法达到了最优的分类性能。

基于DFECANet的遥感图像飞机目标检测方法

针对现有的遥感图像目标检测方法中对小尺寸飞机目标的检测精度不高、特征信息传递不准确、信息交互不充分等问题,提出了一种基于可辨别特征提取和上下文感知的遥感图像飞机目标检测方法。设计了以可辨别特征提取模块为主体的主干网络,用以加强对多尺度飞机目标的特征提取;引入自适应特征增强模块,选择性关注小目标、优化特征信息的传递与信息交互;并设计了特征融合上采样模块对特征图进行上采样操作,用以提升高层语义信息的准确性。在DOTAv1数据集上的检测精度达到了95.2%,相较于YOLOv5s、SCRDet、ASSD等主流算法,飞机目标的检测精度提高了3.7%~18%。此外,该方法的检测速度以及模型参数量分别为147 fps和13.4 M,相较于当前主流算法具备较强的竞争力,满足在遥感背景下对飞机目标的实时检测需求。

基于HRNet的高分辨率遥感影像道路提取方法

高分辨率遥感影像中,传统的道路提取方法存在着精度低、鲁棒性低的问题,提出基于高分辨率网络(high-resolution net, HRNet)实现高分辨率遥感影像道路分割。对HRNet进行改进,将相同分辨率的HRNet子网的输出与输出层结果进行拼接并输入非局部块,两个损失函数Cross-entropy Loss和Dice Loss用来解决道路数据集样本不平衡问题。实验结果表明,改进的HRNet在公开的CHN6-CUG道路数据集上的分割性能与其他方法相比对道路的提取效果更好,在召回率、均交并比和F1分数3个方面分别达到了97.65%、84.91%和97.25%。

Spatial structural characteristics of the Deda ancient landslide in the eastern Tibetan Plateau:Insights from Audio-frequency Magnetotellurics and the Microtremor Survey Method

It is of crucial importance to investigate the spatial structures of ancient landslides in the eastern Tibetan Plateau’s alpine canyons as they could provide valuable insights into the evolutionary history of the landslides and indicate the potential for future reactivation.This study examines the Deda ancient landslide,situated in the Chalong-ranbu fault zone,where creep deformation suggests a complex underground structure.By integrating remote sensing,field surveys,Audio-frequency Magnetotellurics(AMT),and Microtremor Survey Method(MSM)techniques,along with engineering geological drilling for validation,to uncover the landslide’s spatial feature s.The research indicates that a fault is developed in the upper part of the Deda ancient landslide,and the gully divides it into Deda landslide accumulation zoneⅠand Deda landslide accumulation zoneⅡin space.The distinctive geological characteristics detectable by MSM in the shallow subsurface and by AMT in deeper layers.The findings include the identification of two sliding zones in the Deda I landslide,the shallow sliding zone(DD-I-S1)depth is approximately 20 m,and the deep sliding zone(DD-I-S2)depth is 36.2-49.9 m.The sliding zone(DD-Ⅱ-S1)depth of the DedaⅡlandslide is 37.6-43.1 m.A novel MSM-based method for sliding zone identification is proposed,achieving less than 5%discrepancy in depth determination when compared with drilling data.These results provide a valuable reference for the spatial structural analysis of large-deepseated landslides in geologically complex regions like the eastern Tibetan Plateau.

Study on Ecological Change Remote Sensing Monitoring Method Based on Elman Dynamic Recurrent Neural Network

In this paper, Hailin City of Heilongjiang Province, China is taken as the research area. As an important city in Heilongjiang Province, China, the sustainable development of its ecological environment is related to the opening up, economic prosperity and social stability of Northeast China. In this paper, the remote sensing ecological index (RSEI) of Hailin City in recent 20 years was calculated by using Landsat 5/8/9 series satellite images, and the temporal and spatial changes of the ecological environment in Hailin City were further analyzed and the influencing factors were discussed. From 2003 to 2023, the mean value of RSEI in Hailin City decreased and increased, and the ecological environment decreased slightly as a whole. RSEI declined most significantly from 2003 to 2008, and it increased from 2008 to 2013, decreased from 2013 to 2018, and increased from 2018 to 2023 again, with higher RSEI value in the south and lower RSEI value in the northwest. It is suggested to appropriately increase vegetation coverage in the northwest to improve ecological quality. As a result, the predicted value of Elman dynamic recurrent neural network model is consistent with the change trend of the mean value, and the prediction error converges quickly, which can accurately predict the ecological environment quality in the future study area.

改进UNet++的遥感影像森林变化检测方法

针对森林覆盖变化检测任务,现有的基于深度学习模型存在结构复杂且忽视光谱与空间协同关系的问题,导致检测效果并不理想。为了解决这个问题,本研究提出一种结合多尺度空间解耦卷积(MSDConv)和空-谱特征协同策略(SSFC)的改进UNet++轻量级森林覆盖变化检测方法。首先,基于UNet++网络构建一个非权重共享伪孪生网络,增加少量参数便能实现更好的特征提取,采用MSDConv模块捕捉变化对象的多尺度特征,减少信息冗余和参数计算;其次,在MSDConv中引入SSFC,获取空间和谱间的三维注意力权重且不增加额外参数,使得MSDConv获取更丰富的边缘和细节特征;最后,使用6种植被指数增强森林覆盖变化特征。结果表明,本研究提出的模型森林覆盖变化检测精度、召回率和F1分数分别为93.12%,93.62%和93.37%,模型参数量和计算量分别为6.28 MB和11.25 GB。与原始Sami-UNet++方法对比,本研究提出的模型准确率、召回率和F1分数仅分别下降1.41%、1.66%和1.53%,但参数量与计算量分别降低5.76 MB和16.19 GB。本研究提出的模型显著提高了森林覆盖变化检测任务的检测效率,对于需要处理大量图像数据的森林覆盖变化检测任务具有重要的意义,可为森林灾害的评估以及森林资源的保护提供技术手段。

D-Net:一种动态模拟道路形态的路网提取模型

对于当前利用深度学习技术实现遥感影像中道路提取时,由于网络中的矩形卷积而导致无法捕捉道路形态特征、提取精度低等问题。基于此,提出一种基于形变卷积的卷积神经网络(D-Net,Deformation Network),该网络旨在从遥感影像中准确提取道路区域。为了验证模型的性能,在马萨诸塞州道路数据集和DeepGlobe道路数据集上进行了测试。此外,还选择了一景GF-2影像对模型的泛化能力进行评估。实验结果显示,D-Net显著提升了道路分割的精度,为通过形变卷积优化提取结果提供了有力支持。该研究的开展对于进一步探索道路智能提取具有重要的理论和实践指导意义。

基于UNet-ResNet14^(*)半监督学习的无人机影像森林树种分类

无人机遥感在森林树种精细和高效分类制图中具有巨大的潜力。为了快速准确获取森林的优势树种分布信息,该研究探讨了半监督学习方法在树种分类方面的有效性。以福建省福州市、龙岩市和三明市的4个试验区为例,构建精简的ResNet18为主干的UNet树种分类模型(UNet-ResNet14^(*)),使用交叉熵和Dice系数的联合损失函数来优化模型参数,对比分析Self-training和Mean Teacher两种不同的半监督学习方法在无人机影像森林树种分类模型的泛化能力。结果表明,以ResNet14^(*)作为主干的分类模型与其他模型相比精度更高且预测速度更快,当联合损失函数权重值为0.5的情况下模型预测效果最好,总体精度达到了91.15%。经过Self-training的模型在木荷、马尾松、杉木3个样本充足的类别中精度均有所提升,总精度为91.08%,比原始模型略低,但在独立验证区的精度为88.50%,比原始模型高;Mean Teacher方法的总精度为88.56%,在独立验证区的精度为73.56%。因此,研究认为可以采用Self-trainin半监督方法结合UNet-ResNet14^(*)的方案快速得到试验区的树种组成信息。

基于UNet模型的遥感影像建筑物变化检测研究

UNet是一个典型的对称U型网络,针对该网络无法准确地捕捉到建筑物的边界及细节信息的问题,提出了一个改进的UNet网络模型,即在UNet网络模型的跳跃连接中加入可以提高网络感知力的scSE注意力模块,同时将模型中的编码器更换为能够更好地捕捉图像细节和纹理的VGG19,对公开数据集LEVIR-CD进行建筑物变化检测实验。实验结果表明,该方法较于原方法虽然精确率下降0.66%,但是召回率和F1分数分别提高了13.18%和5.17%,说明该方法有效提升了UNet网络模型对建筑物边界及细节的识别,使建筑物变化检测的精度得到有效提升。