Design of Content-Based Retrieval System in Remote Sensing Image Database

To retrieve the object region efficaciously from massive remote sensing image database, a model for content-based retrieval of remote sensing image is given according to the characters of remote sensing image application firstly, and then the algorithm adopted for feature extraction and multidimensional indexing, and relevance feedback by this model are analyzed in detail. Finally, the contents intending to be researched about this model are proposed.

Remote Sensing Image Retrieval Based on 3D-Local Ternary Pattern(LTP)Features and Non-subsampled Shearlet Transform(NSST)Domain Statistical Features

With the increasing popularity of high-resolution remote sensing images,the remote sensing image retrieval(RSIR)has always been a topic of major issue.A combined,global non-subsampled shearlet transform(NSST)-domain statistical features(NSSTds)and local three dimensional local ternary pattern(3D-LTP)features,is proposed for high-resolution remote sensing images.We model the NSST image coefficients of detail subbands using 2-state laplacian mixture(LM)distribution and its three parameters are estimated using Expectation-Maximization(EM)algorithm.We also calculate the statistical parameters such as subband kurtosis and skewness from detail subbands along with mean and standard deviation calculated from approximation subband,and concatenate all of them with the 2-state LM parameters to describe the global features of the image.The various properties of NSST such as multiscale,localization and flexible directional sensitivity make it a suitable choice to provide an effective approximation of an image.In order to extract the dense local features,a new 3D-LTP is proposed where dimension reduction is performed via selection of‘uniform’patterns.The 3D-LTP is calculated from spatial RGB planes of the input image.The proposed inter-channel 3D-LTP not only exploits the local texture information but the color information is captured too.Finally,a fused feature representation(NSSTds-3DLTP)is proposed using new global(NSSTds)and local(3D-LTP)features to enhance the discriminativeness of features.The retrieval performance of proposed NSSTds-3DLTP features are tested on three challenging remote sensing image datasets such as WHU-RS19,Aerial Image Dataset(AID)and PatternNet in terms of mean average precision(MAP),average normalized modified retrieval rank(ANMRR)and precision-recall(P-R)graph.The experimental results are encouraging and the NSSTds-3DLTP features leads to superior retrieval performance compared to many well known existing descriptors such as Gabor RGB,Granulometry,local binary pattern(LBP),Fisher vector(FV),vector of locally aggregated descriptors(VLAD)and median robust extended local binary pattern(MRELBP).For WHU-RS19 dataset,in terms of{MAP,ANMRR},the NSSTds-3DLTP improves upon Gabor RGB,Granulometry,LBP,FV,VLAD and MRELBP descriptors by{41.93%,20.87%},{92.30%,32.68%},{86.14%,31.97%},{18.18%,15.22%},{8.96%,19.60%}and{15.60%,13.26%},respectively.For AID,in terms of{MAP,ANMRR},the NSSTds-3DLTP improves upon Gabor RGB,Granulometry,LBP,FV,VLAD and MRELBP descriptors by{152.60%,22.06%},{226.65%,25.08%},{185.03%,23.33%},{80.06%,12.16%},{50.58%,10.49%}and{62.34%,3.24%},respectively.For PatternNet,the NSSTds-3DLTP respectively improves upon Gabor RGB,Granulometry,LBP,FV,VLAD and MRELBP descriptors by{32.79%,10.34%},{141.30%,24.72%},{17.47%,10.34%},{83.20%,19.07%},{21.56%,3.60%},and{19.30%,0.48%}in terms of{MAP,ANMRR}.The moderate dimensionality of simple NSSTds-3DLTP allows the system to run in real-time.

Spectrum Feature Retrieval and Comparison of Remote Sensing Images Using Improved ISODATA Algorithm

Due to the large quantities of data and high relativity of the spectra of remote sensing images, K-L transformation is used to eliminate the relativity. An improved ISODATA(Interative Self-Organizing Data Analysis Technique A) algorithm is used to extract the spectrum features of the images. The computation is greatly reduced and the dynamic arguments are realized. The comparison of features between two images is carried out, and good results are achieved in simulation.

An improved SVM model for relevance feedback in remote sensing image retrieval

With the rapid development of satellite remote sensing technology and an ever-increasing number of Earth observation satellites being launched,the global volume of remotely sensed imagery has been growing exponentially.Processing the variety of remotely sensed data has increasingly been complex and difficult.It is also hard to efficiently and intelligently retrieve what users need from a massive database of images.This paper introduces an improved support vector machine(SVM)model,which optimizes the model parameters and selects the feature subset based on the particle swarm optimization(PSO)method and genetic algorithm(GA)for remote sensing image retrieval.The results from an image retrieval experiment show that our method outperforms traditional methods such as GRID,PSO,and GA in terms of consistency and stability.

An improved Bag-of-Words framework for remote sensing image retrieval in large-scale image databases

Due to advances in satellite and sensor technology,the number and size of Remote Sensing(RS)images continue to grow at a rapid pace.The continuous stream of sensor data from satellites poses major challenges for the retrieval of relevant information from those satellite datastreams.The Bag-of-Words(BoW)framework is a leading image search approach and has been successfully applied in a broad range of computer vision problems and hence has received much attention from the RS community.However,the recognition performance of a typical BoW framework becomes very poor when the framework is applied to application scenarios where the appearance and texture of images are very similar.In this paper,we propose a simple method to improve recognition performance of a typical BoW framework by representing images with local features extracted from base images.In addition,we propose a similarity measure for RS images by counting the number of same words assigned to images.We compare the performance of these methods with a typical BoW framework.Our experiments show that the proposed method has better recognition performance than that of the BoW and requires less storage space for saving local invariant features.

Web-based remote sensing image retrieval using multiscale and multidirectional analysis based on Contourlet and Haralick texture features

Purpose-Extracting suitable features to represent an image based on its content is a very tedious task.Especially in remote sensing we have high-resolution images with a variety of objects on the Earth’s surface.Mahalanobis distance metric is used to measure the similarity between query and database images.The low distance obtained image is indexed at the top as high relevant information to the query.Design/methodology/approach-This paper aims to develop an automatic feature extraction system for remote sensing image data.Haralick texture features based on Contourlet transform are fused with statistical features extracted from the QuadTree(QT)decomposition are developed as feature set to represent the input data.The extracted features will retrieve similar images from the large image datasets using an image-based query through the web-based user interface.Findings-The developed retrieval system performance has been analyzed using precision and recall and F1 score.The proposed feature vector gives better performance with 0.69 precision for the top 50 relevant retrieved results over other existing multiscale-based feature extraction methods.Originality/value-The main contribution of this paper is developing a texture feature vector in a multiscale domain by combining the Haralick texture properties in the Contourlet domain and Statistical features using QT decomposition.The features required to represent the image is 207 which is very less dimension compare to other texture methods.The performance shows superior than the other state of art methods.

An improved coverage-oriented retrieval algorithm for large-area remote sensing data

With the rapid development of satellite technology,the amount of remote sensing data and demand for remote sensing data analysis over large areas are greatly increasing.Hence,it is necessary to quickly filter out an optimal dataset from massive dataset to support various remote sensing applications.However,with the improvements in temporal and spatial resolution,remote sensing data have become fragmented,which brings challenges to data retrieval.At present,most data service platforms rely on the query engines to retrieve data.Retrieval results still have a large amount of data with a high degree of overlap,which must be manually selected for further processing.This process is very labour-intensive and time-consuming.This paper proposes an improved coverage-oriented retrieval algorithm that aims to retrieve an optimal image combination with the minimum number of images closest to the imaging time of interest while maximized covering the target area.The retrieval efficiency of this algorithm was analysed by applying different implementation practices:Arcpy,PyQGIS,and GeoPandas.The experimental results confirm the effectiveness of the algorithm and suggest that the GeoPandas-based approach is most advantageous when processing large-area data.

面向无人机绝对定位的遥感影像快速检索方法

针对在复杂环境下无人机景象匹配导航中的视觉绝对定位问题,提出了一种聚合深度学习特征的实时影像快速检索方法。首先,引入可训练软分配深度学习框架—NetVLAD,结合VGG16网络提取并聚合生成影像稳定的全局特征表达向量;其次,在初始检索阶段,使用KD树结构对影像全局特征向量构建检索索引,在不损失检索精度的前提下提高检索速度;最后,使用皮尔逊积矩相关系数对初始检索结果进行快速预判断,自动过滤初始检索结果,对于需要重排序的影像则采用特征学习匹配算法——图神经网络SuperGlue进行匹配重排序。所提方法在公开的夏季和冬季遥感影像数据集分组进行实验,实验结果表明:未重排序条件下,初始检索结果第一张影像平均准确率达到了58.27%,部分特征较好地区准确率达到了85%,对不同时相遥感影像也有很好的适应性,平均检索一张影像耗时3.7 s,可为无人机景象匹配导航的初始定位提供参考。

基于注意力机制和软匹配的多标签遥感图像检索方法

针对卷积神经网络对于多标签遥感图像特征提取能力弱、不能准确反映遥感图像多标签复杂性的问题,提出基于注意力机制和软匹配的多标签遥感图像检索方法。在特征提取阶段,以密集卷积神经网络模型为基础,在每个密集块(Dense Block)后添加CBAM(Convolutional Block Attention Module)层,实现对多标签图像区域特征提取。在模型训练时,利用区分硬匹配与软匹配的联合损失函数,学习图像的哈希编码表示。通过评估遥感图像哈希编码间的汉明距离,实现相似图像的检索。实验结果表明,所提方法在数据集NUS-WIDE和多标签遥感图像数据集DLRSD上与其他基于全局特征的深度哈希方法相比,明显提升了检索准确率。

Study on retrieve specified objects in massive remote sensing data

In this paper, we show that to retrieve specified objects in massive remote sensing data set is very important in both practice and theory. An algorithm-based content retrieval in the massive data set is studied. To avoid the loss of information, the algorithm based on the Support Vector Machine classification is proposed. Also, the experiment on the real data set is made.

无序无人机影像的并行化SfM三维重建方法

增量式运动恢复结构(ISfM)已成为无人机影像三维重建的关键技术。然而,大数据量、大重叠度和高分辨率的无序无人机影像导致的匹配对检索代价大、迭代优化误差累积和效率低的问题,使其难以满足大场景ISfM处理需求。本文提出联合全局描述子和图索引的无人机影像并行化SfM方法。针对影像特征数量大、影像检索编码本尺寸增加导致的匹配对检索效率低的问题,设计了联合全局描述子和图索引的高效影像检索方法,从而加速影像匹配。针对分块并行化SfM子场景合并存在同名点搜索效率低、内存消耗大、合并解算精度低的问题,设计了基于按需匹配图和双向重投影误差的子场景合并方法,实现无人机影像的并行化SfM重建。利用不同场景、不同采集方式获取的真实无人机影像进行试验,结果表明本文方法能够实现36~108倍加速比的匹配对检索,ISfM重建效率达到30倍加速,且相对定向和绝对定向精度与传统方法相当。

基于布局化-语义联合表征遥感图文检索方法

遥感图文检索可以从类别繁多、内容复杂的遥感数据中检索到有价值的信息,对环境评估、城市规划以及灾害预测具有重要意义。但是,遥感图文跨模态检索存在一个关键问题,即遥感图像的空间布局信息被忽略。其主要体现在2个方面:①遥感目标的远距离建模困难;②遥感相邻次要目标被淹没。基于以上问题,提出了一种基于布局化-语义联合表征的跨模态遥感图像文本检索(SL-SJR),主要包括主导语义监督的布局化视觉特征提取(DSSL)模块、布局化视觉-全局语义交叉指导(LV-GSCG)模块和多视角匹配(MVM)模块。DSSL模块实现主导语义类别特征监督下图像的布局化建模。LV-GSCG模块计算布局化视觉特征与文本中提取的全局语义特征的相似度来实现不同模态特征的交互。MVM模块建立跨模态特征指导的多视角度量匹配机制以消除跨模态数据之间的语义鸿沟。在4个基线遥感图像文本数据集上的实验验证,结果表明所提方法在大多数跨模态遥感图像文本检索任务中可以达到最先进的性能。

基于组合优化的遥感图文检索轻量化

针对现有基于图网络的遥感图文检索模型存在的海量参数、模型时效性低、存储空间需求大等问题,提出一种基于组合优化的遥感图文检索轻量化方法。从模型架构角度,设计基于跨阶段融合的轻量化卷积模块精简图文检索模型的参数;从数值量化角度,设计图网络混合精度训练与量化推理策略提升模型推理速度。在多个遥感检索数据集上的实验结果表明,该方法在检索精度基本不下降的条件下,总参数量、浮点运算量相比于典型方法降低60%以上。

面向遥感图像检索的无监督哈希融合方法

遥感图像数据规模庞大,且大部分数据没有标注,因此能够独立于数据标注的无监督哈希算法更适用于遥感图像检索。本文提出了一种针对无监督哈希的融合方法,首先,通过随机游走算法得到遥感图像预训练特征之间的流形相似度,并结合特征之间的余弦相似度构造相似指示矩阵,该矩阵可以度量无监督哈希码的有效性。然后,将哈希码的有效性作为节点,哈希码之间的排序相关性作为边来动态地构建关联图,并将图中的连通分量作为哈希码的组合,避免可能产生退化结果的哈希码组合,进而降低计算复杂度。最后,将哈希码的归一化有效性作为权重,对每种组合方案进行自适应的晚期融合,生成判别能力更强的哈希码。在2个数据集上的一系列实验表明,该方法能自适应地选择出合适的融合方案,有效提升融合哈希码的检索性能,并且在付出更小训练代价的情况下,获得接近穷举方法检索性能的融合方案。

Liquid Water Path Retrieval Using the Lowest Frequency Channels of Fengyun-3C Microwave Radiation Imager(MWRI)

The Microwave Radiation Imager （MWRI） on board Chinese Fengyun-3 （FY-3） satellites provides measurements at 10.65, 18.7, 23.8, 36.5, and 89.0 GHz with both horizontal and vertical polarization channels. Brightness temperature measurements of those channels with their central frequencies higher than 19 GHz from satellite-based microwave imager radiometers had traditionally been used to retrieve cloud liquid water path （LWP） over ocean. The results show that the lowest frequency channels are the most appropriate for retrieving LWP when its values are large. Therefore, a modified LWP retrieval algorithm is developed for retrieving LWP of different magnitudes involving not only the high frequency channels but also the lowest frequency channels of FY-3 MWRI. The theoretical estimates of the LWP retrieval errors are between 0.11 and 0.06 mm for 10.65- and 18.7-GHz channels and between 0.02 and 0.04 mm for 36.5- and 89.0-GHz channels. It is also shown that the brightness temperature observations at 10.65 GHz can be utilized to better retrieve the LWP greater than 3 mm in the eyewall region of Super Typhoon Neoguri （2014）. The spiral structure of clouds within and around Typhoon Neoguri can be well captured by combining the LWP retrievals from different frequency channels.

一种顾及空间语义的跨模态遥感影像检索技术

随着遥感影像获取的场景和目标内容日益丰富,传统的基于关键字和属性字段的检索手段无法反映对于影像内容的语义检索,导致用户无法从大规模影像中获取满足需求语义的数据。OpenAI发布的语言-图像预训练对比模型(CLIP),为跨模态开放要素检索提供了重要的模型支撑,但其在顾及空间语义关系等复杂跨模态检索任务上能力不足。本文提出了一种顾及空间语义关系的跨模态遥感影像检索技术,基于CLIP构建跨模态遥感影像检索模型GEOCLIP,通过对比学习方法训练,习得富含空间语义与开放信息的双模态语义对齐公共表示空间,特别针对遥感影像跨模态空间语义检索问题,引入遥感影像和文本表达中的空间关系提取,实现融合空间语义的跨模态检索。本文提出的顾及空间语义的跨模态遥感影像检索技术,在RSICD Dataset数据集上进行了验证,其R@1,R@5,R@10和mR指标均达到目前最优,其中平均召回率mR相较于CLIP提升了3.45%,相较于已公开发表的最优方法GaLR提升了77.22%。GEOCLIP在各种空间查询上的平均召回率mR全部优于CLIP,其中针对at、near、around的空间查询提升效果最大,分别为3.72%、8.85%、7.11%。

基于深度多相似性哈希方法的遥感图像检索

哈希方法由于低存储、高效率的特性而被广泛应用于遥感图像检索领域。面向遥感图像检索任务的无监督哈希方法存在伪标签不可靠、图像对的训练权重相同以及图像检索精度较低等问题,为此,提出一种基于深度多相似性哈希(DMSH)的遥感图像检索方法。针对优化伪标签和训练关注度分别构建自适应伪标签模块(APLM)和成对结构信息模块(PSIM)。APLM采用K最近邻和核相似度来评估图像间的相似关系,实现伪标签的初始生成和在线校正。PSIM将图像对的多尺度结构相似度映射为训练关注度,为其分配不同的训练权重从而优化深度哈希学习。DMSH通过Swin Transformer骨干网络提取图像的高维特征,将基于语义相似矩阵的伪标签作为监督信息以训练深度网络,同时网络在两个基于不同相似度设计的模块上实现交替优化,充分挖掘图像间的多种相似信息进而生成具有高辨识力的哈希编码,实现遥感图像的高精度检索。实验结果表明,DMSH在EuroSAT和PatternNet数据集上的平均精度均值较对比方法分别提高0.8%~3.0%和9.8%~12.5%,其可以在遥感图像检索任务中取得更高的准确率。

基于卷积神经网络的光学遥感影像分析综述

光学遥感影像包含大量的地物信息,图像复杂性高,如何充分利用影像中的特征信息准确进行识别一直是该领域应用的一个难题。卷积神经网络具有对复杂特征进行选择和提取的优势,在遥感影像识别中有着优异的表现。介绍光学遥感影像的特点和经典的卷积神经网络及其在光学遥感影像中的研究实例,并基于遥感影像数据集进行了网络性能分析。从场景分类、目标检测和图像检索三大领域,详细综述常用的遥感影像数据集和研究进展,并作算法性能分析。最后给出基于卷积神经网络的光学遥感影像识别在未来的研究方向。

基于北京二号卫星影像与同步实测数据的椒江入海口水质遥感反演

利用2020年7月23日北京二号卫星影像和现场同步实测水质数据,建立水质参数与影像波段的一元和多元线性回归方程,选择决定系数(R^(2))最高的方程对台州椒江入海口水域的氨氮、总磷和COD_(Mn)进行遥感反演,并对3种水质参数进行精度评价.结果显示,在氨氮、总磷和COD_(Mn) 3种水质参数中,采用4个波段组合的反演模型的R^(2)均为最高.氨氮、总磷和COD_(Mn)的精度检验结果中,均方根误差分别为0.03、0.18和4.24 mg/L,平均相对误差分别为15.86%、24.41%和23.91%,反演精度从高到低依次为氨氮、COD_(Mn)和总磷.3种水质参数水平均从上游向下游递增,最高值都出现在靠近入海口处;地表水环境质量等级也大致呈现从上游至下游水质等级逐渐变差的趋势.

基于机载激光雷达与高景一号数据的草原地上生物量反演研究

草原地上生物量(AGB)是草原调查监测中的重要指标,是草原生态保护和资源合理利用的依据,对草原可持续发展与科学管理具有重要意义。本研究以广西兴安县热性灌草丛为研究对象,结合机载激光雷达数据与高分辨率多光谱卫星影像,利用2021年采集的89个实地样方调查数据,对草原AGB进行了遥感反演研究。结果表明,草层高度信息是草原AGB建模的重要指标。增强型植被指数(EVI)、比值植被指数(RVI)、归一化植被指数(NDVI)中EVI与AGB的相关系数最高(0.666),高度指标中平均草层高度(CHM_(mean))与AGB的相关系数最高(0.686),二者结合的指标中RVI×CHM_(mean)与AGB的相关系数最高(0.735)。模型精度验证结果显示,EVI模型中均方根误差(RMSE)最低,为292.047 g·m^(-2),CHM_(mean)模型中RMSE最低,为245.084 g·m^(-2),RVI×CHM_(mean)模型中RMSE最低为225.872 g·m^(-2)。结果说明机载激光雷达数据可以有效提取草层高度信息,尽管存在明显的低估现象,但在草原AGB研究中仍具有较大的应用潜力。