面向高光谱图像分类的空谱判别分析

针对传统的基于特征提取的高光谱图像地物分类算法大多只考虑光谱信息而忽略空间信息的问题,提出了一种面向高光谱分类的半监督空谱全局与局部判别分析(S3 GLDA)算法。该算法首先利用少量标记样本保存数据集的线性可分性和全局判别信息,再依靠较多的无标记的空间局部近邻像元来揭示局部判别信息和非线性局部流形,使高光谱遥感图像的光谱域全局判别结构和空间域局部判别结构在低维特征空间同时得以保留,并在输出特征中自动融入了空间信息,构成了半监督的空谱判别分析。在Indian Pines和PaviaU数据集的实验表明,总体分类精度分别达到76.24%和82.96%。与现有几种算法比较,该算法有效提高了输出特征在低维空间的判别能力,更好地揭示了数据集的内在非线性多模本质,有效提升了高光谱图像数据集的地物分类精度。     

加权空-谱与最近邻分类器相结合的高光谱图像分类

提出了一种基于加权空-谱距离(WSSD)的相似性度量方法 ,并将其应用到最近邻分类器(KNN)中,导出了一种新的高光谱图像分类算法。该算法利用高光谱图像的物理特性,通过引入空间窗口和光谱因子这两个参数来挖掘出图像中的空间信息与光谱信息,利用空间近邻点对中心像元进行重构。在最大限度减少图像冗余信息的基础上,增大了同类像元间的相似性以及异类像元间的差异性,获得了更为有效的鉴别特征,从而更好地实现了数据间的相似性度量。基于Indian Pines和PaviaU高光谱数据集进行了实验,结果表明:将提出的WSSD-KNN算法应用于高光谱图像分类时,其分类精度高于其他算法,总体分类精度分别达到了91.72%和96.56%。由于算法较好地融合了图像中的空间-光谱信息,提取出了更为有效的鉴别特征,故不仅有效地改善了高光谱数据的地物分类精度,而且可在训练样本较少时,保持较高的识别率。     

空-谱协同流形重构的高光谱影像分类

鉴于传统高光谱影像分类大都采用监督学习方法,且仅利用了光谱信息,未考虑影像空间特征和流形结构。提出一种基于空-谱协同流形重构误差的高光谱影像分类方法,该算法基于高光谱影像中地物分布的空间一致性,利用少量标记的样本和大量的无标记空间近邻样本来进行半监督学习,并利用测试样本在每一子流形上的重构误差来表征相似性,实现鉴别分类。在Indian Pines和University of Pavia数据集上的实验结果表明,本文方法的分类精度在各种条件下要优于其他分类算法,其最高总体精度分别达到了95.67%和91.92%。该算法将高光谱遥感影像中的空间-光谱信息融入不同地物的子流形结构表征,在训练样本数量较少时仍能得到好的分类效果,有效提升了分类性能。     

融合空谱-梯度特征的深度高光谱图像去噪

为了去除高光谱图像采集过程中产生的噪声,提升后续图像处理的性能,提出了一种融合空谱-梯度特征的深度高光谱图像去噪方法.它包括空谱特征网络和梯度特征网络,且各网络使用密集跳跃连接和可分离卷积策略进行优化.空谱网络模型实现噪声特征的精确提取,梯度网络模型对噪声纹理特征进行补充,最后基于两个网络的特征提取结果进行融合,实现噪...     

改进高斯过程回归的高光谱空谱联合分类算法

针对高斯过程回归在高光谱图像分类中计算量较大、分类精度较低等问题,提出一种基于改进高斯过程回归的高光谱空谱联合分类算法。算法以最大方差为指标选取样本的子集缩小高斯过程回归参数求解的计算范围,采用平方根矩阵分解法对新添加样本进行模型结果预测,有效提升运算效率;算法以空间-光谱特征信息为基础,在像元近邻空间中重新定义邻域像元空-谱关联距离,将融入空间近邻信息的空-谱关联距离作为权值来度量邻域像元相似性,加大同类地物归为近邻的概率,从而提高地物分类的精度。在Indian Pines和Pavia University两组高光谱数据集上进行仿真实验,实验结果可知,与其他同类算法横向相比,本文提出的改进算法在总体分类精度、平均分类精度和Kappa系数等评价指标至少提高了2.3%,1.4%和1.07%,与改进前的模型算法纵向对比可知,本文提出的改进算法在取得较高总体分类精度的同时,大幅降低了算法的运行时间。     

空谱联合预测高光谱图像无损压缩rice算法

针对rice算法低维预测不能有效降低高光谱数据冗余问题,提出基于空谱联合预测的低复杂度rice算法,应用于高光谱图像无损压缩.根据高光谱图像三维数据特征建立三维预测模型,利用相邻波段谱间相关系数进行联合预测系数分配,有效地减少了高光谱图像空间和谱间冗余.提出基于预测误差均值的最优编码参数选择算法,计算复杂度由O(N)降为O(1).实验结果表明,本文方法提高无损压缩比5%～40%,编码时间较经典rice算法缩短了4%以上,有利于实时处理和工程实现.     

加权空谱局部保持投影的高光谱图像特征提取

为了提高高光谱图像的分类精度,有效利用高光谱图像的空间信息和光谱信息对高光谱图像进行预处理,本文提出了一种新的空谱联合特征提取方法,加权空-谱局部保持投影算法(WSSLPP)。该算法结合高光谱图像的物理特性对高光谱图像进行重构,降低了图像中奇异点的干扰;然后对局部像素近邻保持嵌入(LPNPE)和局部保持投影(LPP)的目标函数进行加权求和,有效融合高光谱图像空间维和光谱维的信息来构建投影矩阵。WSSLPP不仅保留了高光谱图像在空间维上像素间的近邻关系,而且保持了在光谱维上样本的固有结构,有利于高光谱图像的分类。在Indian Pines和PaviU数据库上对该算法进行验证分析,结果表明:基于WSSLPP算法得到的分类精度明显高于其他算法,总体分类精度的最大值分别为99.00%,99.50%,有效提高了高光谱图像的分类精度。     

高光谱遥感图像空谱联合分类方法研究

在遥感影像研究领域里,高光谱数据分类是一个热点问题。近年来,在这个问题上涌现出很多研究方法,然而,大多数方法都是用浅层的方法提取原始数据的特征。将深度学习的方法引入高光谱图像分类中,提出一种新的基于深信度网络(DBN)的特征提取方法和图像分类架构用于高光谱数据分析。将谱域-空域特征提取和分类器相结合提高分类精度。使用高光谱数据进行实验,结果表明该分类器优于当前的一些先进的分类方法。此外,本文还揭示了深度学习系统在高光谱图像分类研究中具有的巨大潜力。     

基于分数阶微分的高光谱图像特征提取与分类

针对高光谱遥感图像的特征提取与地物分类,提出一种基于分数阶微分的高光谱图像特征提取方法,设计二维分数阶微分掩模提取高光谱图像的像素空间分数阶微分（SpaFD）特征,并提出一种空谱联合准则用于选取微分掩模阶数。为充分利用高光谱图像的空间特征与光谱特征,将SpaFD特征与原始特征直连融合获得SpaFD-Spe-Spa混合特征,并采用三维卷积神经网络（3DCNN）、先采用主成分分析（PCA）对像素光谱进行降维处理再送入三维卷积神经网络（3DCNNPCA）以及采用混合光谱网络（HybridSN）验证SpaFD-Spe-Spa混合特征的有效性。实验中分别采用3×3,5×5和7×7的分数阶微分掩模进行空间特征提取,4个真实高光谱图像的实验结果表明,所提取的SpaFD特征和SpaFDSpe-Spa特征可有效提升高光谱图像的地物分类精度,且SpaFD-Spe-Spa特征对地物分类准确率的提升更为明显：SpaFD特征相比原始特征在Indian Pines,Botswana,Pavia University和Salinas 4个数据上的分类识别率在最优情况下分别提升了3.87%,1.42%,2.41%和2...     

马铃薯晚疫病害的高光谱图像空谱对比研究

为了快速检测马铃薯晚疫病,采用高光谱成像技术对马铃薯晚疫病的空谱信息进行对比研究以得到最佳判别手段。使用高光谱相机采集病害侵染0～6 d的高光谱图像,同时选取第6 d典型晚疫病病害的高光谱数据作为研究对象。采用二阶导数结合主成分分析和二次主成分分析分别从光谱和空间两个方面进行特征提取,之后基于特征波段反射率和主成分图像灰度值建立K最近邻分类算法、BP神经网络、决策树算法3种识别模型对不同时期病害进行识别。实验结果表明:基于二次主成分图像的灰度值结合BP神经网络建立的模型对马铃薯晚疫病的识别具有良好的成效,其识别率达96.6%。利用主成分图像灰度值建立的3种模型既减少了波段的冗余又提高了识别率,为研究和开发实时在线检测仪器提供了参考。     

Dilated and soft attention-guided convolutional neural network for breast cancer histology images classification

Breast cancer is one of the most common types of cancer in women, and histopathological imaging is considered the gold standard for its diagnosis. However, the great complexity of histopathological images and the considerable workload make this work extremely time-consuming, and the results may be affected by the subjectivity of the pathologist. Therefore, the development of an accurate, automated method for analysis of histopathological images is critical to this field. In this article, we propose a deep learning method guided by the attention mechanism for fast and effective classification of haematoxylin and eosin-stained breast biopsy images. First, this method takes advantage of DenseNet and uses the feature map's information. Second, we introduce dilated convolution to produce a larger receptive field. Finally, spatial attention and channel attention are used to guide the extraction of the most useful visual features. With the use of fivefold cross-validation, the best model obtained an accuracy of 96.47% on the BACH2018 dataset. We also evaluated our method on other datasets, and the experimental results demonstrated that our model has reliable performance. This study indicates that our histopathological image classifier with a soft attention-guided deep learning model for breast cancer shows significantly better results than the latest methods. It has great potential as an effective tool for automatic evaluation of digital histopathological microscopic images for computer-aided diagnosis.     

基于深度卷积神经网络的车标分类

基于深度学习框架Caffe和具有强大计算能力的GPU,运用深度神经网络AlexNet和GoogleNet,对具有不同背景的汽车图像进行网络训练,以达到车辆自动识别的目的。分别对4类车标进行网络训练与测试,实验结果表明,在图像识别分类方面,与传统识别方法相比,深度卷积神经网络更具优势。     

循环神经网络多标签航空图像分类

由于航空图像背景复杂,包含的物体类别多样,航空图像分类任务仍然面临困难。针对传统航空图像多标签分类算法准确率低、泛化性差的问题,本文提出了一种基于循环神经网络多标签航空图像分类方法。首先,采用超像素分割获取图像的低层特征,通过注意力机制生成注意力特征图;接着,采用交叉验证的方式获取最佳的图像尺度,将多尺度注意力特征图嵌入卷积神经网络中对图像进行特征提取;最后,采用改进的双向长短期记忆网络挖掘标签之间的相关性,改进的双向长短期记忆网络增加了输入门到输出门之间的连接,使输入状态可以更好地控制每一内存单元输出的信息,并且将遗忘门和输入门合并成单一的更新门,使得改进的双向长短期记忆网络可以学到更长时期的历史信息。结果显示,在图像变换尺度为1,1.3,2时,模型在UCM多标签数据集上的精确率和召回率分别达到了85.33%和87.05%,F1值达到了0.862。本文方法相比于原始VGGNet16模型,精确率提高了7.25%,召回率提高了8.94%。实验表明,该方法可以有效提高航空图像多标签分类任务的准确率。     

基于CNN的火炮身管全景图像疵病识别方法

针对火炮身管内膛疵病种类多、定性定量分析难和检测自动化程度低等问题,本文提出一种以卷积神经网络为基础的疵病识别方法。首先,对全景图像进行预处理,主要包括全景展开、光照强度调整、膛线去除等;其次,通过最优阈值法对图像进行二值化处理,并利用四连通域法提取疵病区域;最后,采用卷积神经网络对疵病进行自动的分类识别。实验结果表明,该方法能有效避免人工疵病特征提取和人工特征描述计算等复杂步骤,实现了"采集-识别-判定"全过程的自动运行,真正实现了窥膛检测的自动化,身管疵病的识别率超过92%,识别准确率远高于基于统计学原理及支持向量机的分类方式,具有较高的准确性,为火炮身管修复及寿命预估等奠定了坚实的基础。     

代价敏感卷积神经网络：一种机械故障数据不平衡分类方法

机械设备实际工作过程中正常样本丰富、故障样本匮乏,卷积神经网络在处理这种分布不平衡的数据时对少数类的识别率很低。为解决上述问题,提出一种代价敏感卷积神经网络,首先经过多层卷积和池化运算学习原始监测数据中的机械设备本征性能状态知识;其次通过全连接层将本征性能状态知识映射为机械设备健康状态;最后利用代价敏感损失函数为少数类样本赋予较大的误分类代价,实现对不平衡的机械故障数据的有效分类。为验证所提方法的有效性,使用具有不同不平衡比的刀具数据集和轴承数据集,利用代价敏感卷积神经网络以及主流的分类算法分别测试其对于不平衡数据的分类性能。实验结果表明,所提方法对不平衡数据集中的少数类样本识别率相对于传统卷积神经网络提升了22%以上。     

Microscopic segmentation and classification of COVID-19 infection with ensemble convolutional neural network

The detection of biological RNA from sputum has a comparatively poor positive rate in the initial/early stages of discovering COVID-19, as per the World Health Organization. It has a different morphological structure as compared to healthy images, manifested by computer tomography (CT). COVID-19 diagnosis at an early stage can aid in the timely cure of patients, lowering the mortality rate. In this reported research, three-phase model is proposed for COVID-19 detection. In Phase I, noise is removed from CT images using a denoise convolutional neural network (DnCNN). In the Phase II, the actual lesion region is segmented from the enhanced CT images by using deeplabv3 and ResNet-18. In Phase III, segmented images are passed to the stack sparse autoencoder (SSAE) deep learning model having two stack auto-encoders (SAE) with the selected hidden layers. The designed SSAE model is based on both SAE and softmax layers for COVID19 classification. The proposed method is evaluated on actual patient data of Pakistan Ordinance Factories and other public benchmark data sets with different scanners/mediums. The proposed method achieved global segmentation accuracy of 0.96 and 0.97 for classification.     

Bayesian model for detection and classification of meningioma nuclei in microscopic images

Image segmentation aims to determine structures of interest inside a digital picture in biomedical sciences. State‐of‐the art automatic methods however still fail to provide the segmentation quality achievable by humans who employ expert knowledge and use software to mark target structures on an image. Manual segmentation is time‐consuming, tedious and suffers from interoperator variability, thus not serving the requirements of daily use well. Therefore, the approach presented here abandons the goal of full‐fledged segmentation and settles for the localization of circular objects in photographs (10 training images and 20 testing images with several hundreds of nuclei each). A fully trainable softcore interaction point process model was hence fit to the most likely locations of nuclei of meningioma cells. The Broad Bioimage Benchmark Collection/SIMCEP data set of virtual cells served as controls. A ‘colour deconvolution’ algorithm was integrated to determine (based on anti‐Ki67 immunohistochemistry) which real cells might have the potential to proliferate. In addition, a density parameter of the underlying Bayesian model was estimated. Immunohistochemistry results were ‘simulated'for the virtual cells. The system yielded true positive (TP) rates in the detection and classification of real nuclei and their virtual counterparts. These hits outnumbered those obtained from the public domain image processing software ImageJ by 10%. The method introduced here can be trained to function not only in medicine and morphology‐based systems biology but in other application domains as well. The algorithm lends itself to an automated approach that constitutes a valuable tool which is easy to use and generates acceptable results quickly.     

Deep convolutional tree-inspired network: a decision-tree-structured neural network for hierarchical fault diagnosis of bearings

The fault diagnosis of bearings is crucial in ensuring the reliability of rotating machinery. Deep neural networks have provided unprecedented opportunities to condition monitoring from a new perspective due to the powerful ability in learning fault-related knowledge. However, the inexplicability and low generalization ability of fault diagnosis models still bar them from the application. To address this issue, this paper explores a decision-tree-structured neural network, that is, the deep convolutional tree-inspired network (DCTN), for the hierarchical fault diagnosis of bearings. The proposed model effectively integrates the advantages of convolutional neural network (CNN) and decision tree methods by rebuilding the output decision layer of CNN according to the hierarchical structural characteristics of the decision tree, which is by no means a simple combination of the two models. The proposed DCTN model has unique advantages in 1) the hierarchical structure that can support more accuracy and comprehensive fault diagnosis, 2) the better interpretability of the model output with hierarchical decision making, and 3) more powerful generalization capabilities for the samples across fault severities. The multiclass fault diagnosis case and cross-severity fault diagnosis case are executed on a multicondition aeronautical bearing test rig. Experimental results can fully demonstrate the feasibility and superiority of the proposed method.