Automated Deep Learning Driven Crop Classification on Hyperspectral Remote Sensing Images

Hyperspectral remote sensing/imaging spectroscopy is a novel approach to reaching a spectrum from all the places of a huge array of spatial places so that several spectral wavelengths are utilized for making coherent images.Hyperspectral remote sensing contains acquisition of digital images from several narrow,contiguous spectral bands throughout the visible,Thermal Infrared(TIR),Near Infrared(NIR),and Mid-Infrared(MIR)regions of the electromagnetic spectrum.In order to the application of agricultural regions,remote sensing approaches are studied and executed to their benefit of continuous and quantitativemonitoring.Particularly,hyperspectral images(HSI)are considered the precise for agriculture as they can offer chemical and physical data on vegetation.With this motivation,this article presents a novel Hurricane Optimization Algorithm with Deep Transfer Learning Driven Crop Classification(HOADTL-CC)model onHyperspectralRemote Sensing Images.The presentedHOADTL-CC model focuses on the identification and categorization of crops on hyperspectral remote sensing images.To accomplish this,the presentedHOADTL-CC model involves the design ofHOAwith capsule network(CapsNet)model for generating a set of useful feature vectors.Besides,Elman neural network(ENN)model is applied to allot proper class labels into the input HSI.Finally,glowworm swarm optimization(GSO)algorithm is exploited to fine tune the ENNparameters involved in this article.The experimental result scrutiny of the HOADTL-CC method can be tested with the help of benchmark dataset and the results are assessed under distinct aspects.Extensive comparative studies stated the enhanced performance of the HOADTL-CC model over recent approaches with maximum accuracy of 99.51%.

Classification of hyperspectral remote sensing images based on simulated annealing genetic algorithm and multiple instance learning

A hybrid feature selection and classification strategy was proposed based on the simulated annealing genetic algonthrn and multiple instance learning （MIL）. The band selection method was proposed from subspace decomposition, which combines the simulated annealing algorithm with the genetic algorithm in choosing different cross-over and mutation probabilities, as well as mutation individuals. Then MIL was combined with image segmentation, clustering and support vector machine algorithms to classify hyperspectral image. The experimental results show that this proposed method can get high classification accuracy of 93.13% at small training samples and the weaknesses of the conventional methods are overcome.

Towards complex scenes: A deep learning-based camouflaged people detection method for snapshot multispectral images

Camouflaged people are extremely expert in actively concealing themselves by effectively utilizing cover and the surrounding environment. Despite advancements in optical detection capabilities through imaging systems, including spectral, polarization, and infrared technologies, there is still a lack of effective real-time method for accurately detecting small-size and high-efficient camouflaged people in complex real-world scenes. Here, this study proposes a snapshot multispectral image-based camouflaged detection model, multispectral YOLO(MS-YOLO), which utilizes the SPD-Conv and Sim AM modules to effectively represent targets and suppress background interference by exploiting the spatial-spectral target information. Besides, the study constructs the first real-shot multispectral camouflaged people dataset(MSCPD), which encompasses diverse scenes, target scales, and attitudes. To minimize information redundancy, MS-YOLO selects an optimal subset of 12 bands with strong feature representation and minimal inter-band correlation as input. Through experiments on the MSCPD, MS-YOLO achieves a mean Average Precision of 94.31% and real-time detection at 65 frames per second, which confirms the effectiveness and efficiency of our method in detecting camouflaged people in various typical desert and forest scenes. Our approach offers valuable support to improve the perception capabilities of unmanned aerial vehicles in detecting enemy forces and rescuing personnel in battlefield.

Image Fusion Based on NSCT and Sparse Representation for Remote Sensing Data

The practice of integrating images from two or more sensors collected from the same area or object is known as image fusion.The goal is to extract more spatial and spectral information from the resulting fused image than from the component images.The images must be fused to improve the spatial and spectral quality of both panchromatic and multispectral images.This study provides a novel picture fusion technique that employs L0 smoothening Filter,Non-subsampled Contour let Transform(NSCT)and Sparse Representation(SR)followed by the Max absolute rule(MAR).The fusion approach is as follows:first,the multispectral and panchromatic images are divided into lower and higher frequency components using the L0 smoothing filter.Then comes the fusion process,which uses an approach that combines NSCT and SR to fuse low frequency components.Similarly,the Max-absolute fusion rule is used to merge high frequency components.Finally,the final image is obtained through the disintegration of fused low and high frequency data.In terms of correlation coefficient,Entropy,spatial frequency,and fusion mutual information,our method outperforms other methods in terms of image quality enhancement and visual evaluation.

Mapping of Aluminous Rich Laterite Depositions through Hyper Spectral Remote Sensing

Increased dimensionality of the satellite data proves to be very useful for discriminating features with very close spectral matching. Present study concentrates on the retrieval of reflectance spectra from the level one radiometrically corrected data in Koraput district (Orissa) for the Bauxite ore. In the present study, atmospheric correction model FLAASH has been used to retrieve reflectance from the radiance data. Preprocessing of the dataset has been done before applying atmospheric correction on the dataset. Spectral subsetting of noise prone bands has been successfully done. Local destriping of the affected bands has been done using a 3*3 local mean filter. Spectral signatures of samples were derived from the processed data. Spectral signature of each sample and derived features vectors were correlated with the satellite image of the area and distribution of each feature was demarcated. Spatial abundance of each feature was used in preparation of mineral abundance map. Accuracy of the map was assessed using training sets of representative geological units. The mineral abundance mapping using the spectral analysis of the reflectance image involves the endmember collection using the N-Dimensional visualizer tool in ENVI software. Laterite, Bauxite, Iron and silica rich Aluminous laterite soil, Alluvium and Forest were selected as the end members after understanding the geology and analysis of the reflectance image. Various mapping techniques were applied to generate the final classified mineral abundance Map, Linear Spectral Unmixing, Mixture Tune Matched Filtering, Spectral Feature Fitting, Spectral Angle Mapper were the techniques used. Results have revealed the ability of Hyper spectral Remote sensing data for the identification and mapping of Hydrothermal altered products like Bauxite, Aluminous Laterite. This technology can be utilized for targeting minerals in the altered zone.

Class-guided coupled dictionary learning for multispectral-hyperspectral remote sensing image collaborative classification

The fine classification of large-scale scenes is becoming more and more important in optical remote sensing applications.As two kinds of typical optical remote sensing data,multispectral images(MSIs)and hyperspectral images(HSIs)have complementary characteristics.The MSI has a large swath and short revisit period,but the number of bands is limited with low spectral resolution,leading to weak separability of between class spectra.Compared with MSI,HSI has hundreds of bands and each of them is narrow in bandwidth,which enable it to have the ability of fine classification,but too long in aspects of revisit period.To make efficient use of their combined advantages,multispectral-hyperspectral remote sensing image collaborative classification has become one of hot topics in remote sensing.To deal with the collaborative classification,most of current methods are unsupervised and only consider the HSI reconstruction as the objective.In this paper,a class-guided coupled dictionary learning method is proposed,which is obviously distinguished from the current methods.Specifically,the proposed method utilizes the labels of training samples to construct discriminative sparse representation coefficient error and classification error as regularization terms,so as to enforce the learned coupled dictionaries to be both representational and discriminative.The learned coupled dictionaries facilitate pixels from the same category have similar sparse represent coefficients,while pixels from different categories have different sparse represent coefficients.The experiments on three pairs of HSI and MSI have shown better classification performance.

Early detection of pine wilt disease in Pinus tabuliformis in North China using a field portable spectrometer and UAV-based hyperspectral imagery

Background:Pine wilt disease(PWD)is a major ecological concern in China that has caused severe damage to millions of Chinese pines(Pinus tabulaeformis).To control the spread of PWD,it is necessary to develop an effective approach to detect its presence in the early stage of infection.One potential solution is the use of Unmanned Airborne Vehicle(UAV)based hyperspectral images(HIs).UAV-based HIs have high spatial and spectral resolution and can gather data rapidly,potentially enabling the effective monitoring of large forests.Despite this,few studies examine the feasibility of HI data use in assessing the stage and severity of PWD infection in Chinese pine.Method:To fill this gap,we used a Random Forest(RF)algorithm to estimate the stage of PWD infection of trees sampled using UAV-based HI data and ground-based data(data directly collected from trees in the field).We compared relative accuracy of each of these data collection methods.We built our RF model using vegetation indices(VIs),red edge parameters(REPs),moisture indices(MIs),and their combination.Results:We report several key results.For ground data,the model that combined all parameters(OA:80.17%,Kappa:0.73)performed better than VIs(OA:75.21%,Kappa:0.66),REPs(OA:79.34%,Kappa:0.67),and MIs(OA:74.38%,Kappa:0.65)in predicting the PWD stage of individual pine tree infection.REPs had the highest accuracy(OA:80.33%,Kappa:0.58)in distinguishing trees at the early stage of PWD from healthy trees.UAV-based HI data yielded similar results:the model combined VIs,REPs and MIs(OA:74.38%,Kappa:0.66)exhibited the highest accuracy in estimating the PWD stage of sampled trees,and REPs performed best in distinguishing healthy trees from trees at early stage of PWD(OA:71.67%,Kappa:0.40).Conclusion:Overall,our results confirm the validity of using HI data to identify pine trees infected with PWD in its early stage,although its accuracy must be improved before widespread use is practical.We also show UAV-based data PWD classifications are less accurate but comparable to those of ground-based data.We believe that these results can be used to improve preventative measures in the control of PWD.

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.

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.

Robust Core Tensor Dictionary Learning with Modified Gaussian Mixture Model for Multispectral Image Restoration

The multispectral remote sensing image(MS-RSI)is degraded existing multi-spectral camera due to various hardware limitations.In this paper,we propose a novel core tensor dictionary learning approach with the robust modified Gaussian mixture model for MS-RSI restoration.First,the multispectral patch is modeled by three-order tensor and high-order singular value decomposition is applied to the tensor.Then the task of MS-RSI restoration is formulated as a minimum sparse core tensor estimation problem.To improve the accuracy of core tensor coding,the core tensor estimation based on the robust modified Gaussian mixture model is introduced into the proposed model by exploiting the sparse distribution prior in image.When applied to MS-RSI restoration,our experimental results have shown that the proposed algorithm can better reconstruct the sharpness of the image textures and can outperform several existing state-of-the-art multispectral image restoration methods in both subjective image quality and visual perception.

Multispectral imaging systems for airborne remote sensing to support agricultural production management

This study investigated three different types of multispectral imaging systems for airborne remote sensing to support management in agricultural application and production.The three systems have been used in agricultural studies.They range from low-cost to relatively high-cost,manually operated to automated,multispectral composite imaging with a single camera and integrated imaging with custom-mounting of separate cameras.Practical issues regarding use of the imaging systems were described and discussed.The low-cost system,due to band saturation,slow imaging speed and poor image quality,is more preferable to slower moving platforms that can fly close to the ground,such as unmanned autonomous helicopters,but not recommended for low or high altitude aerial remote sensing on fixed-wing aircraft.With the restriction on payload unmanned autonomous helicopters are not recommended for high-cost systems because they are typically heavy and difficult to mount.The system with intermediate cost works well for low altitude aerial remote sensing on fixed-wing aircraft with field shapefile-based global positioning triggering.This system also works well for high altitude aerial remote sensing on fixed-wing aircraft with global positioning triggering or manually operated.The custom-built system is recommended for high altitude aerial remote sensing on fixed-wing aircraft with waypoint global positioning triggering or manually operated.

Advances in spaceborne hyperspectral remote sensing in China

With the maturation of satellite technology,Hyperspectral Remote Sensing(HRS)platforms have developed from the initial ground-based and airborne platforms into spaceborne platforms,which greatly promotes the civil application of HRS imagery in the fields of agriculture,forestry,and environmental monitoring.China is playing an important role in this evolution,especially in recent years,with the successful launch and operation of a series of civil hyper-spectral spacecraft and satellites,including the Shenzhou-3 spacecraft,the Gaofen-5 satellite,the SPARK satellite,the Zhuhai-1 satellite network for environmental and resources monitoring,the FengYun series of satellites for meteorological observation,and the Chang’E series of spacecraft for planetary exploration.The Chinese spaceborne HRS platforms have various new characteristics,such as the wide swath width,high spatial resolution,wide spectral range,hyperspectral satellite networks,and microsatellites.This paper focuses on the recent progress in Chinese spaceborne HRS,from the aspects of the typical satellite systems,the data processing,and the applications.In addition,the future development trends of HRS in China are also discussed and analyzed.

Hyperspectral Reflectance Characteristics of Cyanobacteria

Potentially harmful cyanobacterial blooms are an emerging environmental concern in freshwater bodies worldwide. Cyanobacterial blooms are generally caused by high nutrient inputs and warm, still waters and have been appearing with increasing frequency in water bodies used for drinking water supply and recreation, a problem which will likely worsen with a warming climate. Cyanobacterial blooms are composed of genera with known biological pigments and can be distinguished and analyzed via hyperspectral image collection technology such as remote sensing by satellites, airplanes, and drones. Here, we utilize hyperspectral microscopy and imaging spectroscopy to characterize and differentiate several important bloom-forming cyanobacteria genera obtained in the field during active research programs conducted by US Geological Survey and from commercial sources. Many of the cyanobacteria genera showed differences in their spectra that may be used to identify and predict their occurrence, including peaks and valleys in spectral reflectance. Because certain cyanobacteria, such as Cylindrospermum or Dolichospermum, are more prone to produce cyanotoxins than others, the ability to differentiate these species may help target high priority waterbodies for sampling. These spectra may also be used to prioritize restoration and research efforts to control cyanobacterial harmful algal blooms (CyanoHABs) and improve water quality for aquatic life and humans alike.

Detecting Oil Spill Contamination Using Airborne Hyperspectral Data in the River Nile, Egypt

Egypt is a highly populated country of about 85 million inhabitants that are concentrated on the Nile Delta and on the flood plain of the Nile River. More than 90% of this population relies on the Nile River in their water demand for domestic use. Currently, Egypt is facing a problem with the trans-boundary water budget coming from the Nile basin. This urges for managing the water quantity and quality to secure the water needs. This paper discusses the potential use of airborne hyperspectral data for water quality management in the form of detecting the oil contamination in the Nile River in integration with in-situ measurements including ASD spectroradiometer and eco-sounder multi-probe devices. The eco-sounder multi-probe device measured most of the water quality parameters and detected the existence of oil contamination at 1200 bb downstream of the study area. The airborne hyperspectral images were analyzed and calibrated with the spectral library determined from the in-situ spectroradiometer to map the patches of the oil contamination. The details of the findings and learning lessons are fully discussed in the paper.

Fusion of Remote Sensing Images Based on Nonsubsampled Contourlet Transform and Region Segmentation

The purpose of remote sensing images fusion is to produce a fused image that contains more clear,accurate and comprehensive information than any single image.A novel fusion method is proposed in this paper based on nonsubsampled contourlet transform(NSCT) and region segmentation.Firstly,the multispectral image is transformed to intensity-hue-saturation(IHS) system.Secondly,the panchromatic image and the component intensity of the multispectral image are decomposed by NSCT.Then the NSCT coefficients of high and low frequency subbands are fused by different rules,respectively.For the high frequency subbands,the fusion rules are also unalike in the smooth and edge regions.The two regions are segregated in the panchromatic image,and the segmentation is based on particle swarm optimization.Finally,the fusion image can be obtained by performing inverse NSCT and inverse IHS transform.The experimental results are evaluated by both subjective and objective criteria.It is shown that the proposed method can obtain superior results to others.

Fusion of Landsat 8 OLI and PlanetScope Images for Urban Forest Management in Baton Rouge, Louisiana

In recent years image fusion method has been used widely in different studies to improve spatial resolution of multispectral images. This study aims to fuse high resolution satellite imagery with low multispectral imagery in order to assist policymakers in the effective planning and management of urban forest ecosystem in Baton Rouge. To accomplish these objectives, Landsat 8 and PlanetScope satellite images were acquired from United States Geological Survey (USGS) Earth Explorer and Planet websites with pixel resolution of 30m and 3m respectively. The reference images (observed Landsat 8 and PlanetScope imagery) were acquired on 06/08/2020 and 11/19/2020. The image processing was performed in ArcMap and used 6-5-4 band combination for Landsat 8 to visually inspect healthy vegetation and the green spaces. The near-infrared (NIR) panchromatic band for PlanetScope was merged with Landsat 8 image using the Create Pan-Sharpened raster tool in ArcMap and applied the Intensity-Hue-Saturation (IHS) method. In addition, location of urban forestry parks in the study area was picked using the handheld GPS and recorded in an excel sheet. This sheet was converted into Excel (.csv) file and imported into ESRI ArcMap to identify the spatial distribution of the green spaces in East Baton Rouge parish. Results show fused images have better contrast and improve visualization of spatial features than non-fused images. For example, roads, trees, buildings appear sharper, easily discernible, and less pixelated compared to the Landsat 8 image in the fused image. The paper concludes by outlining policy recommendations in the form of sequential measurement of urban forest over time to help track changes and allows for better informed policy and decision making with respect to urban forest management.

无人机多光谱影像的小麦倒伏信息多特征融合检测研究

为探究多特征融合方法在作物倒伏领域快速精准识别中的适用性,利用无人机获取多田块冠层尺度的不同倒伏率麦田多光谱数据,对原始倒伏图像进行图像拼接、辐射校正、几何校正等预处理,并利用重归一化差值植被指数和阴影指数分别剔除土壤和阴影背景,提取小麦倒伏DSM模型和植被指数分别与多光谱图像进行多特征图像主成分变换融合,筛选差异性较大的纹理特征,采用支持向量机(SVM)、人工神经网络(ANN)和最大似然法(MLC)监督分类模型对多光谱和DSM融合图像、多光谱和归一化植被指数(NDVI)融合图像、多光谱图像和纹理特征图像进行监督分类,并采用总体精度(OA)、 Kappa系数和提取误差综合评价各监督模型的分类性能和倒伏提取精度。分类结果表明:各监督分类方法在不同倒伏区域提取结果建模效果趋势一致,SVM和ANN整体提取精度高于MLC,在高倒伏区域,多光谱与NDVI融合图像的SVM监督模型(OA:92.63%, Kappa系数:0.85,提取误差:1.11%)提取效果最好;在中倒伏区域,多光谱与DSM融合图像的SVM监督模型(OA:90.35%, Kappa系数:0.79,提取误差:9.34%)提取效果最好;在低倒伏区域,均值纹理特征图像的ANN监督模型(OA:91.05%, Kappa系数:0.82,提取误差:8.20%)提取结果较好。本研究将DSM模型、植被指数、纹理特征与多光谱图像进行融合对比,并对多特征融合方法能否高精度有效提取小麦倒伏信息进行了探究,结果表明无人机多光谱遥感结合特征融合技术能有效提取小麦倒伏面积,提取效果优于单特征小麦倒伏图像。本研究结果可为助力小麦倒伏灾情调查数据的精确获取方法提供参考。

基于时间序列植被指数的小麦条锈病抗性等级鉴定方法

条锈病严重影响小麦产量,培育抗条锈病的小麦品种至关重要。针对传统育种中抗性鉴定手段单一、效率低的问题,该研究提出了一种通过小麦冠层植被指数的时间序列实现对条锈病不同抗性等级的高效鉴定方法。该方法利用无人机采集自然发病的育种群体小麦(共600个样本,516个基因型)冠层多时相的光谱图像,使用随机蛙跳算法和ReliefF算法筛选出6个条锈病病害严重度的敏感特征:归一化色素叶绿素指数(normalized pigment chlorophyll index,NPCI)、沃尔贝克指数(woebbecke index,WI)、叶绿素红边指数(chlorophyll index rededge,CIrededge)、绿大气抵抗植被指数(green atmospherically resistant index,GARI)、归一化差分植被指数(normalized difference vi,NDVI)、叶绿素绿指数(chlorophyll index green,CIgreen),这些敏感特征在试验群体中的时间序列符合条锈病的发病规律,验证了其作为条锈病发病严重度敏感特征的有效性;基于支持向量机(support vector machine,SVM)算法使用上述敏感特征建立条锈病病害严重度等级分类模型,在测试集的表现中,与使用未经过筛选的原始特征所建立的模型相比在精度、平均准确率、平均召回率和F1分数上分别仅下降6.2%、3.3%、2.7%、4.0%,证明了所筛选敏感特征的有效性;针对一般机器学习算法难以捕捉不同抗性等级样本之间较小的特征变化差异的问题,提出了一种从植被指数时间序列转化生成的二维图像中提取特征实现条锈病抗性等级分类的方法。将敏感特征中能够较好区分不同抗病等级的4个时间序列植被指数(NPCI、GARI、NDVI、WI),通过格拉姆角场方法生成格拉姆角和场图像,并制作成数据集,使用DenseNet121网络进行训练,以实现不同条锈病抗病等级的分类。建立的条锈病抗性等级分类模型中,由NPCI时间序列图像建立的分类模型测试效果最佳,其准确率为0.837,召回率为0.834,F1分数可达0.833,能够较好地实现对群体小麦不同品种(系)的条锈病抗性等级差异的区分,表明基于光谱植被指数时间序列的小麦条锈病抗性等级识别方法可以用于小麦抗病育种中抗性等级的鉴定,并可为其他作物的病害抗性等级鉴定提供一定的参考。

一种粒子群优化脉冲耦合神经网络的全色锐化算法

为了进一步降低多光谱与全色影像融合后的光谱和空间信息失真,提高融合质量,文章提出一种粒子群优化脉冲耦合神经网络的多光谱与全色影像融合算法。该算法基于主成分分析和非下采样剪切波搭建融合方法的基础融合框架,在低频系数融合过程中使用细节注射的融合方法,降低非必要的信息注射,从而提高光谱保持度。在融合高频系数时,采用参数自适应的简化脉冲耦合神经网络计算融合权重,并基于粒子群优化算法全局搜索能够获取最佳融合质量的对应参数,以提高空间信息的完整性和清晰度。文章通过三组实验验证提出算法的可行性,并与现有的、经典的融合算法进行对比,实验显示:文章提出的融合算法在三组实验中的光谱角映射均在0.1左右,通用图像质量指数在0.9以上。实验结果表明:该算法不仅能够有效提高全色与多光谱影像的融合质量,而且融合效果稳健,在对比实验中具有最佳的融合性能。

基于改进CNN的猕猴桃根区土壤含水率反演方法

为解决无人机遥感领域根据冠层光谱信息对猕猴桃果树根系土壤含水率(root soil water content, RSWC)进行反演时,现有算法对冠层图像信息分析不足的问题,该研究对传统卷积神经网络模型进行改进,提出一种复合视觉卷积回归神经网络(compound visual convolutional regression network, CVCRNet),该网络复合两种不同尺寸卷积层对图像数据进行卷积特征提取,并使用全连接层对卷积特征值进行降维,从而直接以多光谱图像为分析对象对RSWC进行反演,充分利用多光谱图像内所有数据,提升反演精度。研究采集徐香猕猴桃果树果实膨大期(5-9月)冠层多光谱信息和深度40 cm处的RSWC,把基于图像的CVCRNet网络反演方法与基于植被指数的传统反演方法进行对比,CVCRNet训练结果在验证集R^(2)为0.827,RMSE为0.787%,相较于传统方法在验证集R^(2)为0.759,RMSE为0.983%,反演结果相关性有了明显提升,准确率也有得到一定提高。结果表明,改进后的CNN网络能够作为冠层信息反演的重要工具,在冠层复杂的场景下达成良好的土壤数据反演效果。