多尺度超像素分割和奇异谱分析的高光谱影像分类

目的 高光谱影像（hyperspectral image,HSI）中“同物异谱,异物同谱”的现象普遍存在,使分类结果存在严重的椒盐噪声问题。HSI中的空间地物结构复杂多样,单一尺度的空间特征提取方法无法有效地表达地物类间差异和区分地物边界。有效解决光谱混淆和空间尺度问题是提高分类精度的关键。方法 结合多尺度超像素和奇异谱分析,提出一种新的高光谱影像分类方法,从而充分挖掘地物的局部空间特征和光谱特征,解决空间尺度和光谱混淆的问题,提高分类精度。利用多尺度超像素对影像进行分割,获取不同尺度的分割影像,同时在分割区域内进行均值滤波,减少类内的光谱差异,增强类间的光谱差异;对每个区域计算平均光谱向量,并利用奇异谱分析方法获取光谱的主要鉴别特征,同时消除噪声的影响;利用支持向量机对不同尺度超像素分割影像进行分类,并进行决策融合,得到最终的分类结果。结果 实验选取了两个标准高光谱数据集和一个真实数据集,结果表明,利用本文算法提取的光谱—空间特征进行分类,比直接在原始数据上进行分类分别提高约26.8%、9.2%和13%的精度;与先进的深度学习SSRN （spectral-spatial residual network）算法相比,本文算法在精度上分别提升约5.2%、0.7%和4%,并且运行时间仅为前者的18.3%、45.4%和62.1%,处理效率更高。此外,在训练样本有限的情况下,两个标准数据集的样本分别为1%和0.2%时,本文算法均能取得87%以上的分类精度。结论 针对高光谱影像分类中的难题,提出一种新的融合光谱和多尺度空间特征的HSI分类方法。实验结果表明,本文方法优于对比方法,可以产生更精细的分类结果。     

波段自适应细节注入的高分五号与Sentinel-2遥感影像空谱融合

目的 针对当前空谱融合方法应用到高光谱图像融合时,出现的空间细节信息提升明显但光谱失真,或者光谱保真度高但空间细节信息提升不足的问题,本文提出一种波段自适应细节注入的高分五号（GF-5）高光谱图像（30 m）与Sentinel-2多光谱图像（10 m）的遥感影像空谱融合方法。方法 首先,为了解决两个多波段图像不便于直接融合的问题,提出一种波段自适应的融合策略,对多光谱图像波谱范围以外的高光谱图像波段,以相关系数为标准将待融合图像进行分组。其次,针对传统Gram-Schmidt （GS）融合方法用平均权重系数模拟低分辨率图像造成的光谱失真问题,使用最小均方误差估计计算线性拟合系数,再将拟合图像作为第1分量进行GS正变换,提升融合图像的光谱保真度。最后,为了能同时注入更多的空间细节信息,通过非下采样轮廓波变换将拟合图像、空间细节信息图像和多光谱图像的空间、光谱信息融入到重构的高空间分辨率图像中,再将其与其他GS分量一起进行逆变换,最终得到10 m分辨率的GF-5融合图像。结果 通过与当前用于高光谱图像空谱融合的典型方法比较,本文方法对于受时相影响较小的城镇区域,在提升空间分辨率的同时有较好的光谱保真度,且不会出现噪点;对于受时相变化影响大的植被密集区域,本文方法融合图像有较好的清晰度和地物细节信息,且没有噪点出现。本文方法的CC （correlation coefficient）、ERGAS （erreur relative globale adimensionnelle de synthèse）和SAM （spectral angle mapper）相比于传统GS方法分别提升8%、26%和28%,表明本文方法的光谱保真度大大提高。结论 本文方法的结果空间上没有噪点且光谱曲线与原始光谱曲线基本保持一致,是一种兼具高空间分辨率和高光谱保真度的高光谱图像融合方法。     

Analysis of various optimizers on deep convolutional neural network model in the application of hyperspectral remote sensing image classification

ABSTRACT

Hyperspectral image (HSI) classification is a most challenging task in hyperspectral remote sensing field due to unique characteristics of HSI data. It consists of huge number of bands with strong correlations in the spectral and spatial domains. Moreover, limited training samples make it more challenging. To address such problems, we have presented here a spatial feature extraction technique using deep convolutional neural network (CNN) for HSI classification. As optimizer plays an important role in learning process of deep CNN model, we have presented the effect of seven different optimizers on our deep CNN model in the application of HSI classification. The seven different optimizers used in this study are SGD, Adagrad, Adadelta, RMSprop, Adam, AdaMax, and Nadam. Extensive experimental results on four hyperspectral remote sensing data sets have been presented which demonstrate the superiority of the presented deep CNN model with Adam optimizer for HSI classification.     

高光谱图像空谱特征提取综述

由于高光谱图像包含了丰富的光谱、空间和辐射信息，且具有光谱接近连续、图谱合一的特性，可用于地质勘探、精细农业、生态环境、城市遥感以及军事目标检测等领域的目标精准分类与识别。对高光谱图像进行空谱特征提取是遥感领域的研究热点和前沿课题之一。传统空谱特征提取方法对高光谱图像分类的计算量和样本需求小、理论可解释性好、抗噪声能力强，但应用于分类的精度受限于特征来源；基于深度学习的高光谱图像空谱特征提取方法虽然计算量和样本需求大，但是由于深层空谱特征的表达能力更好，可以大幅度提高分类器的性能。为了便于对高光谱图像空谱特征提取领域进行更深入有效的探索，本文系统综述了相关研究进展。首先，概述了空间纹理与形态学特征提取、空间邻域信息获取及空间信息后处理等传统高光谱空谱特征提取方法的原理，对大量的已有工作进行了梳理、分析与总结。然后，从深度空谱特征提取角度出发，介绍了当前流行的卷积神经网络、图卷积神经网络及跨场景多源数据模型的结构特点及研究进展，分析、评价了基于深度学习的网络模型对高光谱图像空谱特征提取的优势及问题所在。最后，对该研究领域的未来相关发展提出建议并进行了展望。     

高光谱图像超分辨率重建技术研究进展

不同于传统图像（如灰度图像、RGB图像等）专注于保存目标场景的空间信息,高光谱图像蕴含丰富的空—谱信息,不仅可以保存目标的空间信息,还可以保存具有高可辨性的光谱信息。因此高光谱图像广泛应用于多种计算机视觉和遥感图像任务中,如目标检测、场景分类和目标追踪等。然而,在高光谱图像获取以及重建过程中仍然存在许多问题与瓶颈。如传统高光谱成像仪器在成像过程中通常会引入噪声,且获得的图像往往具有较低的空间分辨率,极大地影响了高光谱图像的质量,对后续数据分析任务造成了极大的困难。近年来,高光谱图像超分辨率重建技术研究得到了极大的发展,现有超分辨率重建方法可以大致分为两类,一类为空间超分辨率重建方法,可以通过直接提升高光谱图像的空间分辨率来获得高质量高光谱图像;另一类为光谱超分辨率重建方法,可以通过提升高空间分辨率图像的光谱分辨率来生成高质量高光谱图像。本文从高光谱图像超分辨率重建领域的新设计、新方法和应用场景出发,通过综合国内外前沿文献来梳理该领域的主要发展,重点论述高光谱图像超分辨率重建领域的发展现状、前沿动态、热点问题及趋势。     

基于加权K近邻和卷积神经网络的高光谱图像分类

传统的谱空联合分类算法通常定义一个邻域空间作为空间信息,忽略空间中非邻域空间信息,且容易将异类像元也考虑在内。针对于高光谱图像分类问题,提出了一种加权K近邻算法能够自适应地提取空间信息,首先定义光谱和空间坐标组成的特征空间,利用该特征空间寻找目标像元的K个相似像元,并对这些像元根据特征空间进行加权;将加权后的像元按照一定方式组合成三维张量表示最终的谱空联合信息,使用三维卷积神经网络对其进行训练,得到最终分类结果。从实验结果来看,相对于改进前的算法,在总体分类精度上得到了一定的提升,与原始的三维卷积神经网络相比,在收敛速度上也得到大大提升,为高光谱图像的谱空联合分类提供了一种更加实用的方法。     

利用提升树模型综合Gabor和LPQ特征进行遥感地物识别

为充分融合高光谱遥感图像空间域和频率域的特征信息,提出了一种综合多尺度Gabor和LPQ特征的空谱融合遥感地物识别模型（Ms_GLPQ）。首先,在空间域上利用Gabor滤波器组,提取出遥感图像各类地物多尺度、多方向的空间邻域特征信息,以描述图像的边缘和纹理等空间结构信息;其次,在频率域上将局部相位量化（Local Phase Quantization,LPQ）算子应用于高光谱遥感图像,提取出高光谱图像的多尺度频域纹理特征,获得图像的相位不变特征描述;然后针对其中特征冗余的问题采用主成分分析（PCA）算法进行降维,再将空间域、频率域的特征进行特征融合,获得了能充分描述图像信息的特征向量;最后采用基于提升树的机器学习分类器（XGBoost、CatBoost等）进行识别。在Indian Pines、Salinas和茶树等高光谱遥感数据集上进行学习与分类测试,准确率分别为85.88%、94.42%和92.61%。实验结果表明：与传统方法相比,Ms_GLPQ模型能够提取小比例样本图像中的有效特征,取得了区分性更强的多特征区域描述子,且在采用提升树模型进行分类时效果更优,得到了比常用分类器更高的识别精度。     

用于多光谱和高光谱图像融合的联合自注意力Transformer

目的 将高光谱图像和多光谱图像进行融合，可以获得具有高空间分辨率和高光谱分辨率的光谱图像，提升光谱图像的质量。现有的基于深度学习的融合方法虽然表现良好，但缺乏对多源图像特征中光谱和空间长距离依赖关系的联合探索。为有效利用图像的光谱相关性和空间相似性，提出一种联合自注意力的Transformer网络来实现多光谱和高光谱图像融合超分辨。方法 首先利用联合自注意力模块，通过光谱注意力机制提取高光谱图像的光谱相关性特征，通过空间注意力机制提取多光谱图像的空间相似性特征，将获得的联合相似性特征用于指导高光谱图像和多光谱图像的融合；随后，将得到的融合特征输入到基于滑动窗口的残差Transformer深度网络中，探索融合特征的长距离依赖信息，学习深度先验融合知识；最后，特征通过卷积层映射为高空间分辨率的高光谱图像。结果 在CAVE和Harvard光谱数据集上分别进行了不同采样倍率下的实验，实验结果表明，与对比方法相比，本文方法从定量指标和视觉效果上，都取得了更好的效果。本文方法相较于性能第二的方法EDBIN （enhanced deep blind iterative network），在CAVE数据集上峰值信噪比提高了0.5 dB，在Harvard数据集上峰值信噪比提高了0.6 dB。结论 本文方法能够更好地融合光谱信息和空间信息，显著提升高光谱融合超分图像的质量。     

3D密集全卷积的高光谱地物分类算法研究

针对高光谱遥感图像训练样本较少、光谱维度较高、空间特征与频谱特征存在差异性而导致高光谱地物分类的特征提取不合理、分类精度不稳定和训练时间长等问题,提出了基于3D密集全卷积（3D-DSFCN）的高光谱图像（HSI）分类算法。算法通过密集模块中的3D卷积核分别提取光谱特征和空间特征,采用特征映射模块替换传统网络中的池化层和全连接层,最后通过softmax分类器进行分类。实验结果表明,基于3D-DSFCN的HSI分类方法提高了地物分类的准确率、增强了低频标签的分类稳定性。     

融合多维度卷积神经网络的肺结节分类方法

针对CT图像肺结节分类任务中分类精度低,假阳性高的问题,提出了一种加权融合多维度卷积神经网络的肺结节分类模型,该模型包含两个子模型：基于二维图像的多尺度密集卷积网络模型,以捕获更宽泛的结节变化特征并促进特征重用;基于三维图像的三维卷积神经网络模型,以充分利用结节空间上下文信息。使用二维和三维CT图像训练子模型,根据子模型分类误差计算其权重,对子模型分类结果进行加权融合,得到最终分类结果。该模型在公共数据集LIDC-IDRI上分类准确率达到94.25%,AUC值达到98%。实验结果表明,加权融合多维度模型可以有效地提升肺结节分类性能。     

基于全卷积网络和自编码的高光谱图像分类

针对高光谱图像空间信息利用不足、标记样本数量较少的问题,提出一种基于全卷积网络和堆栈稀疏自编码的高光谱图像分类算法.基于迁移学习的思想,利用预训练好的全卷积网络FCN-8s,挖掘图像潜在的多尺度几何结构特征;选取其特征的像素邻域信息,采用拼接融合的方法与原光谱信息进行融合;利用堆栈稀疏自编码网络完成最终的多尺度空谱特征...     

结合Contourlet和HSI变换的组合优化遥感图像融合方法

针对遥感图像中全色图像与多光谱图像融合问题,提出一种组合优化图像融合方法——COFM.通过HIS变换获取多光谱图像的亮度分量后,采用Contourlet变换对全色图像和多光谱图像的亮度分量进行分解,分别获取其高频和低频子图;对高频子图提取分形特征,采用取最大的融合规则进行融合;对低频子图提取能量特征后采用第二代非支配排序遗传算法选择融合权值;然后使用加权模型对其进行融合.实验结果表明,COFM的融合效果优于传统图像融合方法,能够在提升图像空间分辨率的同时较好地保留光谱信息.     

Multi-attentive hierarchical dense fusion net for fusion classification of hyperspectral and LiDAR data

With recent advance in Earth Observation techniques, the availability of multi-sensor data acquired in the same geographical area has been increasing greatly, which makes it possible to jointly depict the underlying land-cover phenomenon using different sensor data. In this paper, a novel multi-attentive hierarchical fusion net (MAHiDFNet) is proposed to realize the feature-level fusion and classification of hyperspectral image (HSI) with Light Detection and Ranging (LiDAR) data. More specifically, a triple branch HSI-LiDAR Convolutional Neural Network (CNN) backbone is first developed to simultaneously extract the spatial features, spectral features and elevation features of the land-cover objects. On this basis, hierarchical fusion strategy is adopted to fuse the oriented feature embeddings. In the shallow feature fusion stage, we propose a novel modality attention (MA) module to generate the modality integrated features. By fully considering the correlation and heterogeneity between different sensor data, feature interaction and integration is released by the proposed MA module. At the same time, self-attention modules are also adopted to highlight the modality specific features. In the deep feature fusion stage, the obtained modality specific features and modality integrated features are fused to construct the hierarchical feature fusion framework. Experiments on three real HSI-LiDAR datasets demonstrate the effectiveness of the proposed framework. The code will be public on https://github.com/SYFYN0317/-MAHiDFNet.     

Tripartite-structure transformer for hyperspectral image classification

Hyperspectral images contain rich spatial and spectral information, which provides a strong basis for distinguishing different land-cover objects. Therefore, hyperspectral image (HSI) classification has been a hot research topic. With the advent of deep learning, convolutional neural networks (CNNs) have become a popular method for hyperspectral image classification. However, convolutional neural network (CNN) has strong local feature extraction ability but cannot deal with long-distance dependence well. Vision Transformer (ViT) is a recent development that can address this limitation, but it is not effective in extracting local features and has low computational efficiency. To overcome these drawbacks, we propose a hybrid classification network that combines the strengths of both CNN and ViT, names Spatial-Spectral Former(SSF). The shallow layer employs 3D convolution to extract local features and reduce data dimensions. The deep layer employs a spectral-spatial transformer module for global feature extraction and information enhancement in spectral and spatial dimensions. Our proposed model achieves promising results on widely used public HSI datasets compared to other deep learning methods, including CNN, ViT, and hybrid models.     

Research on fundus image registration and fusion method based on nonsubsampled contourlet and adaptive pulse coupled neural network

We present a registration and fusion method of fluorescein fundus angiography image and color fundus image which combines Nonsubsampled Contourlet (NSCT) and adaptive Pulse Coupled Neural Network (PCNN). Firstly, we register two images by Speeded Up Robust Features (SURF) feature points, the nearest neighbor and the next nearest neighbor distance ratio method to eliminate the spatial difference between the source images. Secondly, we use Random Sample Consensus (RANSAC) algorithm to achieve precise matching of feature points. Then, according to the transformation parameters obtained by RANSAC algorithm, we perform spatial transformation on the floating image to complete the registration. Finally, we obtain the low-frequency sub-band and high-frequency sub-band of the image to be fused by NSCT decomposition. The low-frequency sub-band is fused by the regional energy. The high-frequency sub-bands are studied using a simplified-PCNN model and the Particle Swarm Optimization algorithm. The link strength of the simplified-PCNN is an improved Laplacian energy and the images are fused based on the number of times the pixels are ignited. The proposed method has higher average gradient (AG) value and information entropy (IE) value and lower relative global dimensional synthesis error (ERGAS) than the existing fusion methods of the fundus image. The fusion image can accurately synthesize the image information, clarify the performance of the details, and has better spectral quality in the spectral range. The image of fused provides an effective reference for the clinical diagnosis of fundus diseases.

     

基于空谱融合网络的高光谱图像分类方法

针对高光谱图像分类中提取的空-谱特征表达能力弱及维数较高的问题,提出一种基于空-谱融合网络（SSF-Net）的高光谱图像分类方法。首先,利用双通道卷积神经网络（Two-CNN）同时提取高光谱图像的光谱和空间特征;其次,使用多模态压缩双线性池化（MCB）将所提取的多模态特征向量的外积投射到低维空间,以此产生空-谱联合特征。该特征融合网络,既可以分析光谱特征和空间特征向量中元素之间的复杂关系,同时也避免对光谱和空间向量直接进行外积计算,造成维数过高、计算困难的问题。最终实验表明,与现有基于神经网络的分类方法相比,所提出的高光谱图像分类算法能够获得更高的像元分类精度,表明该网络所提取的空-谱联合向量对高光谱图像具有更强的特征表达能力。     

Spectral-spatial K-Nearest Neighbor approach for hyperspectral image classification

Hyperspectral image (HSI) classification is a very active research topic in remote sensing and has numerous potential applications. This paper presents a simple but effective classification method based on spectral-spatial information and K-nearest neighbor (KNN). To be specific, we propose a spectral-spatial KNN (SSKNN) method to deal with the HSI classification problem, which effectively exploits the distances all neighboring pixels of a given test pixel and training samples. In the proposed SSKNN framework, a set-to-point distance is exploited based on least squares and a weighted KNN method is used to achieve stable performance. By using two standard HSI benchmark, we evaluate the proposed method by comparing it with eight competing methods. Both qualitative and quantitative results demonstrate our SSKNN method achieves better performance than other ones.     

A new residual fusion classification method for hyperspectral images

In this paper, we propose a novel residual fusion classification method for hyperspectral image using spatial–spectral information, abbreviated as RFC-SS. The RFC-SS method first uses the Gabor texture features and the non-parametric weighted spectral features to describe the hyperspectral image from both aspects of spatial and spectral information. Then it applies the residual fusion method to save the useful information from different classification methods, which can greatly improve the classification performance. Finally, the test sample is assigned to the class that has the minimal fused residuals. The RFC-SS classification method is tested on two classical hyperspectral images (i.e. Indian Pines, Pavia University). The theoretical analysis and experimental results demonstrate that the RFC-SS classification method can achieve a better performance in terms of overall accuracy, average accuracy, and the Kappa coefficient when compared to the other classification methods.     

A multiscale feature fusion approach for classification of very high resolution satellite imagery based on wavelet transform

A novel methodology based on multiscale spectral and spatial information fusion using wavelet transform is proposed in order to classify very high resolution (VHR) satellite imagery. Conventional wavelet‐based feature extraction methods employ single windows of a fixed size, which are not satisfactory as the VHR imagery contains complex and multiscale objects. In this paper, spectral and spatial features are extracted based on a set of concentric windows around a central pixel in order to integrate the information across different windows/scales. The proposed method is made up of three blocks: (1) the conventional wavelet‐based feature extraction methods are extended from single band processing to multispectral bands, and from single window to multi‐windows, (2) two multiscale fusion algorithms are proposed to exploit the multiscale spectral and spatial information and (3) a support vector machine (SVM), a relatively new method of machine learning, is used to classify the multiscale spectral–spatial feature sets. The proposed classification method is evaluated on two VHR datasets and the results show that the multiscale approach can improve the classification accuracy in homogeneous areas while simultaneously preserving accuracy in edge regions.