Applications of Hyperspectral Remote Sensing in Ground Object Identification and Classification

Hyperspectral remote sensing has become one of the research frontiers in ground object identification and classification. On the basis of reviewing the application of hyperspectral remote sensing in identification and classification of ground objects at home and abroad. The research results of identification and classification of forest tree species, grassland and urban land features were summarized. Then the researches of classification methods were summarized. Finally the prospects of hyperspectral remote sensing in ground object identification and classification were prospected.

Object-oriented crop classification based on UAV remote sensing imagery

UAV remote sensing images have the advantages of high spatial resolution,fast speed,strong real-time performance,and convenient operation,etc.,and have become a recently developed,vital means of acquiring surface information.It is an important research task for precision agriculture to make full use of the spectrum,texture,color and other characteristic information of crops,especially the spatial arrangement and structure information of features,to explore effective methods for the classification of multiple varieties of crops.In order to explore the applicability of the object-oriented method to achieve accurate classification of UAV high-resolution images,the paper used the object-oriented classification method in ENVI to classify the UAV high-resolution remote sensing image obtained from the orderly structured 28 species of crops in the test field,which mainly includes image segmentation and object classification.The results showed that the plots obtained after classification were continuous and complete,basically in line with the actual situation,and the overall accuracy of crop classification was 91.73%,with Kappa coefficient of 0.87.Compared with the crop planting area based on remote sensing interpretation and field survey,the area error of 17 species of crops in this study was controlled within 15%,which provides a basis for object-oriented crop classification of UAV remote sensing images.

A two-scale approach for estimating forest aboveground biomass with optical remote sensing images in a subtropical forest of Nepal

Forests account for 80%of the total carbon exchange between the atmosphere and terrestrial ecosystems.Thus,to better manage our responses to global warming,it is important to monitor and assess forest aboveground carbon and forest aboveground biomass(FAGB).Different levels of detail are needed to estimate FAGB at local,regional and national scales.Multi-scale remote sensing analysis from high,medium and coarse spatial resolution data,along with field sampling,is one approach often used.However,the methods developed are still time consuming,expensive,and inconvenient for systematic monitoring,especially for developing countries,as they require vast numbers of field samples for upscaling.Here,we recommend a convenient two-scale approach to estimate FAGB that was tested in our study sites.The study was conducted in the Chitwan district of Nepal using GeoEye-1(0.5 m),Landsat(30 m)and Google Earth very high resolution(GEVHR)Quickbird(0.65 m)images.For the local scale(Kayerkhola watershed),tree crowns of the area were delineated by the object-based image analysis technique on GeoEye images.An overall accuracy of 83%was obtained in the delineation of tree canopy cover(TCC)per plot.A TCC vs.FAGB model was developed based on the TCC estimations from GeoEye and FAGB measurements from field sample plots.A coefficient of determination(R2)of 0.76 was obtained in the modelling,and a value of 0.83 was obtained in the validation of the model.To upscale FAGB to the entire district,open source GEVHR images were used as virtual field plots.We delineated their TCC values and then calculated FAGB based on a TCC versus FAGB model.Using the multivariate adaptive regression splines machine learning algorithm,we developed a model from the relationship between the FAGB of GEVHR virtual plots with predictor parameters from Landsat 8 bands and vegetation indices.The model was then used to extrapolate FAGB to the entire district.This approach considerably reduced the need for field data and commercial very high resolution imagery while achieving two-scale forest information and FAGB estimates at high resolution(30 m)and accuracy(R2=0.76 and 0.7)with minimal error(RMSE=64 and 38 tons ha-1)at local and regional scales.This methodology is a promising technique for cost-effective FAGB and carbon estimations and can be replicated with limited resources and time.The method is especially applicable for developing countries that have low budgets for carbon estimations,and it is also applicable to the Reducing Emissions from Deforestation and Forest Degradation(REDD?)monitoring reporting and verification processes.

ART Based Reliable Method for Prediction of Agricultural Land Changes Using Remote Sensing

This paper focuses on prediction of change in agricultural lands by using ART2 algorithm. The existing method used ENVI and ARCGIS software to predict the changes in land, which showed less accuracy due to human errors. To overcome this user friendly GUI based ART2 algorithm has been developed in java to obtain more accuracy in prediction of changes in land. The input is satellite temporal images of the years 1990 and 2014. By using the ART2 algorithm, the input images of the years 1990 and 2014 are classified, where the features are identified to form cluster. The clustered image is given as input and pixel to pixel comparison method in ART2 is implemented in java, for detecting the changes in agricultural lands. The comparison results of ENVI and ARCGIS and GUI based ART2 with in situ data show that the prediction of changes in agricultural land is more accurate in the case of GUI based ART2 implementation.

Classifications of Satellite Imagery for Identifying Urban Area Structures

This study compares three types of classifications of satellite data to identify the most suitable for making city maps in a semi-arid region. The source of our data was GeoEye 1 satellite. To classify this data, two pro-grammes were used: an Object-Based Classification and a Pixel-Based Classification. The second classification programme was further subdi-vided into two groups. The first group included classes (buildings, streets, vacant land, vegetations) which were treated simultaneously and on a single image basis. The second, however, was where each class was identified individually, and the results of each class produced a single image and were later enhanced. The classification results were then as-sessed and compared before and after enhancement using visual then automatic assessment. The results of the evaluation showed that the pix-el-based individual classification of each class was rated the highest after enhancement, increasing the Overall Classification Accuracy by 2%, from 89% to 91.00%. The results of this classification type were adopted for mapping Jeddah’s buildings, roads, and vegetations.

Object-oriented land cover classification using HJ-1 remote sensing imagery

The object-oriented information extraction technique was used to improve classification accuracy,and addressed the problem that HJ-1 CCD remote sensing images have only four spectral bands with moderate spatial resolution.We used two key techniques:the selection of optimum image segmentation scale and the development of an appropriate object-oriented information extraction strategy.With the principle of minimizing merge cost of merging neighboring pixels/objects,we used spatial autocorrelation index Moran's I and the variance index to select the optimum segmentation scale.The Nearest Neighborhood(NN) classifier based on sampling and a knowledge-based fuzzy classifier were used in the object-oriented information extraction strategy.In this classification step,feature optimization was used to improve information extraction accuracy using reduced data dimension.These two techniques were applied to land cover information extraction for Shanghai city using a HJ-1 CCD image.Results indicate that the information extraction accuracy of the object-oriented method was much higher than that of the pixel-based method.

Classification of hyperspectral remote sensing images using frequency spectrum similarity

An algorithm of hyperspectral remote sensing images classification is proposed based on the frequency spectrum of spectral signature.The spectral signature of each pixel in the hyperspectral image is taken as a discrete signal,and the frequency spectrum is obtained using discrete Fourier transform.The discrepancy of frequency spectrum between ground objects' spectral signatures is visible,thus the difference between frequency spectra of reference and target spectral signature is used to measure the spectral similarity.Canberra distance is introduced to increase the contribution from higher frequency components.Then,the number of harmonics involved in the proposed algorithm is determined after analyzing the frequency spectrum energy cumulative distribution function of ground object.In order to evaluate the performance of the proposed algorithm,two hyperspectral remote sensing images are adopted as experimental data.The proposed algorithm is compared with spectral angle mapper (SAM),spectral information divergence (SID) and Euclidean distance (ED) using the product accuracy,user accuracy,overall accuracy,average accuracy and Kappa coefficient.The results show that the proposed algorithm can be applied to hyperspectral image classification effectively.

Land cover classification of remote sensing imagery based on interval-valued data fuzzy c-means algorithm

There is a certain degree of ambiguity associated with remote sensing as a means of performing earth observations.Using interval-valued data to describe clustering prototype features may be more suitable for handling the fuzzy nature of remote sensing data,which is caused by the uncertainty and heterogeneity in the surface spectral reflectance of ground objects.After constructing a multi-spectral interval-valued model of source data and defining a distance measure to achieve the maximum dissimilarity between intervals,an interval-valued fuzzy c-means(FCM)clustering algorithm that considers both the functional characteristics of fuzzy clustering algorithms and the interregional features of ground object spectral reflectance was applied in this study.Such a process can significantly improve the clustering effect;specifically,the process can reduce the synonym spectrum phenomenon and the misclassification caused by the overlap of spectral features between classes of clustering results.Clustering analysis experiments aimed at land cover classification using remote sensing imagery from the SPOT-5 satellite sensor for the Pearl River Delta region,China,and the TM sensor for Yushu,Qinghai,China,were conducted,as well as experiments involving the conventional FCM algorithm,the results of which were used for comparative analysis.Next,a supervised classification method was used to validate the clustering results.The final results indicate that the proposed interval-valued FCM clustering is more effective than the conventional FCM clustering method for land cover classification using multi-spectral remote sensing imagery.

Hyperspectral remote sensing images terrain classification in DCT SRDA subspace

Hyperspectral remote sensing images terrain classification faces the problems of high data dimensionality and lack of labeled training data, resulting in unsatisfied terrain classification efficiency. The feature extraction is required before terrain classification for preserving discriminative information and reducing data dimensionality. A hyperspectral remote sensing images feature extraction method, i.e., discrete cosine transform （DCT） spectral regression discriminant analysis （SRDA） subspace method, was presented to solve the above problems. The proposed DCT SRDA subspace method firstly takes DCT in the original spectral space and gets the DCT coefficients of each pixel spectral curve; secondly performs SRDA in the DCT coefficients space and obtains the DCT SRDA subspace. Minimum distance classifier was designed in the resulting DCT SRDA subspace to evaluate the feature extraction performance. Experiments for two real airborne visible/infrared imaging spectrometer （AVIRIS） hyperspectral images show that, comparing with spectral LDA subspace method, the proposed DCT SRDA subspace method can improve terrain classification efficiency.

Crop Classification Using MODIS NDVI Data Denoised by Wavelet: A Case Study in Hebei Plain, China

Time-series Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data have been widely used for large area crop mapping.However,the temporal crop signatures generated from these data were always accompanied by noise.In this study,a denoising method combined with Time series Inverse Distance Weighted (T-IDW) interpolating and Discrete Wavelet Transform (DWT) was presented.The detail crop planting patterns in Hebei Plain,China were classified using denoised time-series MODIS NDVI data at 250 m resolution.The denoising approach improved original MODIS NDVI product significantly in several periods,which may affect the accuracy of classification.The MODIS NDVI-derived crop map of the Hebei Plain achieved satisfactory classification accuracies through validation with field observation,statistical data and high resolution image.The field investigation accuracy was 85% at pixel level.At county-level,for winter wheat,there is relatively more significant correlation between the estimated area derived from satellite data with noise reduction and the statistical area (R2 = 0.814,p < 0.01).Moreover,the MODIS-derived crop patterns were highly consistent with the map generated by high resolution Landsat image in the same period.The overall accuracy achieved 91.01%.The results indicate that the method combining T-IDW and DWT can provide a gain in time-series MODIS NDVI data noise reduction and crop classification.

Chlorophyll Content Retrieval of Rice Canopy with Multi-spectral Inversion Based on LS-SVR Algorithm

To monitor growth and predict the yield of rice over a large area, the chlorophyll contents in the rice canopy were estimated using the unmanned aerial vehicle(UAV) remote sensing technology. In this work, multi-spectral image information of the rice crop was obtained using a 6-channel multi-spectral camera mounted on a fixed wing UAV, which was flown 600 m above the ground, between 11: 00-14: 00 on a sunny day in summer. The measured chlorophyll values were collected as sample sets. The s-REP index was screened out to estimate chlorophyll contents through the analysis of six kinds of spectral indexes of chlorophyll estimated capacity. An inversion model of the chlorophyll contents was then built using the least square support vector regression(LS-SVR)algorithm, with calibration and prediction R-square values of 0.89 and 0.83, respectively. Finally, remote sensing mapping for a UAV image of the Fangzheng County Dexter Rice Planting Park was accomplished using the inversion model. The inversion and measured values were then compared using regression fitting. R-square and root-mean-square error of the fitting model were 0.79 and 2.39,respectively. The results demonstrated that accurate estimation of rice-canopy chlorophyll contents was feasible using the LS-SVR inversion model developed using the s-REP vegetation index.

Developing an Automated Land Cover Classifier Using LiDAR and High Resolution Aerial Imagery

The aim of this project is to create high resolution land cover classification as well as tree canopy density maps at a regional level using high resolution spatial data. Modeling and the data manipulation and analysis of LiDAR LAS point cloud dataset as well as multispectral aerial photographs from the National Agriculture Imagery Program (NAIP) were carried out. Using geoprocessing modeling, a land cover map is created based on filtered returns from LiDAR point cloud data (LAS dataset) to extract features based on their class and return values, and traditional classification methods of high resolution multi-spectral aerial photographs of the remaining ground cover for Clarion County in Pennsylvania. The newly developed model produced 7 classes at 10 ft × 10 ft spatial resolution, namely: water bodies, structures, streets and paved surfaces, bare ground, grassland, trees, and artificial surfaces (e.g. turf). The model was tested against areas with different sizes (townships and municipalities) which revealed a classification accuracy between 94% and 96%. A visual observation of the results shows that some tree-covered areas were misclassified as built up/structures due to the nature of the available LiDAR data, an area of improvement for further studies. Furthermore, a geoprocessing service was created in order to disseminate the results of the land cover classification as well as the tree canopy density calculation to a broader audience. The service was tested and delivered in the form of a web application where users can select an area of interest and the model produces the land cover and/or the tree canopy density results (http://maps.clarion.edu/LandCoverExtractor). The produced output can be printed as a final map layout with the highlighted area of interest and its corresponding legend. The interface also allows the download of the results of an area of interest for further investigation and/or analysis.

An examination of thematic research,development,and trends in remote sensing applied to conservation agriculture

Conservation agriculture seeks to reduce environmental degradation through sustainable management of agricultural land.Since the 1990s,agricultural research has been conducted using remote sensing technologies;however,few previous reviews have been conducted focused on different conservation management practices.Most of the previous literature has focused on the application of remote sensing in agriculture without focusing exclusively on conservation practices,with some only providing a narrative review,others using biophysical remote sensing for quantitative estimates of the bio-geo-chemical-physical properties of soils and crops,and few others focused on single agricultural management practices.This paper used the preferred reporting items for systematic review(PRISMA)methodology to examine the last 30 years of thematic research,development,and trends associated with remote sensing technologies and methods applied to conservation agriculture research at various spatial and temporal scales.A set of predefined key concepts and keywords were applied in three databases:Scopus,Web of Science,and Google Scholar.A total of 188 articles were compiled for initial examination,where 68 articles were selected for final analysis and grouped into cover crops,crop residue,crop rotation,mulching,and tillage practices.Publications on conservation agriculture research using remote sensing have been increasing since 1991 and peaked at 10 publications in 2020.Among the 68 articles,94%used a pixel-based,while only 6%used an object-based classification method.Prior to 2005,tillage practices were abundantly studied,then crop residue was a focused theme between 2004 and 2012.From 2012 to 2020,the focus shifted again to cover crops.Ten spectral indices were used in 76%of the 68 studies.This examination offered a summary of the new potential and identifies crucial future research needs and directions that could improve the contribution of remote sensing to the provision of long-term operational services for various conservation agriculture applications.

三维卷积与Transformer支持下联合空谱特征的高光谱影像分类

由于CNN对局部特征提取能力强,目前仍是高光谱影像处理和分析中的主流深度模型,但是CNN感受野有限,无法建立长距离依赖关系,学习全局语义信息受限。Transformer的自注意力机制可以对输入序列中的每个位置进行注意力计算,从而能有效获取全局上下文信息。如何实现CNN和Transformer的技术耦合并充分利用空间信息和光谱信息进行高光谱遥感影像分类是一个重要的待研问题。鉴于此,提出一种新的基于三维卷积和Transformer的高光谱遥感影像分类方法,尝试联合空谱特征实现解译能力的提升。使用主成分分析方法对高光谱遥感影像沿垂直方向降维;用非负矩阵分解算法对降维后遥感影像沿水平方向进行空间特征提取,将两种工具处理后遥感影像进行拼接,以充分保留信息;再用三维卷积核对拼接后遥感影像进行空间特征和光谱特征的综合提取;用Transformer的注意力机制对提取空间信息和光谱信息的遥感影像序列建立长距离依赖关系并使用多层感知机完成分类任务。实验表明,所提方法在WHU-Hi龙口、汉川、洪湖以及雄安新区马蹄湾村数据集上均表现出比对比方法更优异的分类性能,表明该方法具有一定的泛化性和稳健性。

Anomalous behaviour detection using one-class support vector machine and remote sensing images: a case study of algal bloom occurrence in inland waters

Algal blooms are a frequent subject in scientific discussions and are the focus of many recent studies,mainly due to their adverse effect on society.Given the lack of ground truth data and the need to develop tools for their detection and monitoring,this research proposes a novel method to automate detection.Concepts derived from multi-temporal image series processing,spectral indices and classification with Oneclass Support Vector Machine(OC-SVM)are used in this proposal.Imagery from multi-spectral sensors on Landsat-8 and MODIS were acquired through the Google Earth Engine API(GEE API).In order to evaluate our method,two bloom detection case studies(Lake Erie(USA)and Lake Taihu(China))were performed.Comparisons were made with methods based on spectral index thresholds.Also,to demonstrate the performance of the OC-SVM classifier compared to other machine learning methods,the proposal was adapted to be used with a Random Forest(RF)classifier,having its results added to the analysis.In situ measurements show that the proposed method delivers highly accurate results compared to spectral index thresholding approaches.However,a drawback of the proposal refers to its higher computational cost.The application of the new method to a real-world bloom case is demonstrated.

基于Sentinel-2多光谱遥感影像的小浪底水质反演

多光谱遥感技术可根据遥感波段信息反演水质参数,降低监测成本,提高监测速度和质量,为大范围水环境监测提供了一种新的方法。通过分析小浪底水库的Sentinel-2多光谱影像以及采样点实测水质数据,建立了最佳光谱波段的水质参数反演模型,对小浪底水库的化学需氧量(COD)、总磷(TP)、总氮(TN)和氨氮(NH_3-N)进行了遥感反演,验证了反演模型的精确度和稳定性,并反演了各水质参数的空间分布规律。结果表明:在4种水质参数反演模型中,COD模型精确度和稳定性最高,其次是TP、TN,最低的是NH_3-N,水库出水口和部分边缘COD质量浓度较高,水库中心TN、TP和NH_3-N质量浓度高于边缘处。

联合运用多光谱和激光雷达技术构建的林分生物量估算模型

以广东省广州市从化石门国家森林公园为研究区域,选择4种不同林分类型(针阔混交林、阔叶林、针叶林、竹林),各林分类型选择3个20 m×20 m地块作为样方;结合激光雷达、多光谱图像、实测数据,构建多元非线性反演模型和多元线性回归模型,估算森林地上生物量,并选择最佳模型进行精度评价。结果表明:(1)依据多源数据建立的4种不同林分类型的多元非线性地上生物量反演模型的精度最高,针阔混交林样地地上生物量预测值为42.79 t·hm^(-2)、阔叶林样地地上生物量预测值为60.46 t·hm^(-2)、针叶林样地地上生物量预测值为32.99 t·hm^(-2)、竹林样地地上生物量预测值为1.92 t·hm^(-2)。(2)研究区中4种不同林分类型的多元非线性地上生物量反演模型的拟合精度,由大到小依次为竹林(决定系数为0.919)、阔叶林(决定系数为0.813)、针叶林(决定系数为0.786)、针阔混交林(决定系数为0.713),均符合精度要求。采用多光谱和激光雷达数据结合的方式,能够较精准地提取林分地上生物量信息,可准确估算针阔混交林、阔叶林、针叶林、竹林的地上生物量。

面向多源异质遥感影像地物分类的自监督预训练方法

近年来,深度学习改变了遥感图像处理的方法。由于标注高质量样本费时费力,标签样本数量不足的现实问题会严重影响深层神经网络模型的性能。为解决这一突出矛盾,本文提出了用于多源异质遥感影像地物分类的自监督预训练和微调分类方案,旨在缓解模型对于标签样本的严重依赖。具体来讲,生成式自监督学习模型由非对称的编码器-解码器结构组成,其中深度编码器从多源遥感数据中学习高阶关键特征,任务特定的解码器用于重建原始遥感影像。为提升特性表示能力,交叉注意力机制模型用于融合异源特征中的信息,进而从多源异质遥感影像中学习更多的互补信息。在微调分类阶段,预训练好的编码器作为无监督特征提取器,基于Transformer结构的轻量级分类器将学习到的特征与光谱信息结合并用于地物分类。这种自监督预训练方案能够从多源异质遥感影像中学习到刻画原始数据的高级关键特征,并且此过程不需要任何人工标注信息,从而缓解了对标签样本的依赖。与现有的分类范式相比,本文提出的自监督预训练和微调方案在多源遥感影像地物分类中能够取得更优的分类结果。

基于迁移学习和多尺度融合的遥感影像场景分类方法研究

随着计算机运算能力的提升以及深度学习技术的发展,无须人工参与的深度学习方法已成为遥感影像分类的主流方法。因此,提出一种基于深度学习并嵌入注意力机制和融合多尺度特征的神经网络对遥感影像进行场景分类。该模型使用迁移学习减少训练样本不足带来的负面影响;在网络中嵌入注意机制、融合多尺度特征来提高对小尺寸地物目标分类的能力,并验证了模型的有效性。通过实验分析得出所提模型对遥感影像场景分类是可行且有效的。

多源多特征集成的南水北调工程丹江库区湿地时空格局演变

丹江口水库是中国南水北调工程的关键水源区。随着城镇化发展以及大坝二次建设,库区湿地生态系统变化显著,亟需湿地生态科学监测。以丹江库区为例,依托Google Earth Engine(GEE)平台,首先采用已有土地覆盖数据集生成湿地样本集,其次整合Landsat影像、DEM等数据构建多源特征集合,并基于Jeffries-Matusita(JM)距离进行特征优选,使用随机森林(RF)算法实现了1985~2023年丹江库区湿地制图。结果表明:①本文提出的样本采集流程可有效提高样本质量,为长时序分类样本采集提供参考;②湿地分类特征优选后特征数由37个减为27个,分类总体精度变化不大,优选后的特征应用于丹江库区湿地分类的平均总体精度(OA)以及平均数量和分配分歧指数(QADI)分别为89.53%和0.0802,说明特征优选有效减少信息冗余,提高影像分类效率;③1985~2023年,丹江库区湿地面积呈波动增加趋势,从1985年的17839.85 ha扩大到2023年的28872.48 ha,面积增长38.12%。总体来说,丹江库区湿地生态系统呈现出逐步恢复和优化的良好态势。