Multivariate Image Analysis in Gaussian Multi-Scale Space for Defect Detection

Inspired by the coarse-to-fine visual perception process of human vision system,a new approach based on Gaussian multi-scale space for defect detection of industrial products was proposed.By selecting different scale parameters of the Gaussian kernel,the multi-scale representation of the original image data could be obtained and used to constitute the multi- variate image,in which each channel could represent a perceptual observation of the original image from different scales.The Multivariate Image Analysis (MIA) techniques were used to extract defect features information.The MIA combined Principal Component Analysis (PCA) to obtain the principal component scores of the multivariate test image.The Q-statistic image, derived from the residuals after the extraction of the first principal component score and noise,could be used to efficiently reveal the surface defects with an appropriate threshold value decided by training images.Experimental results show that the proposed method performs better than the gray histogram-based method.It has less sensitivity to the inhomogeneous of illumination,and has more robustness and reliability of defect detection with lower pseudo reject rate.

STUDY ON THE MULTIVARIATE STATISTICAL ESTIMATION OF TROPICAL CYCLONE INTENSITY USING FY-3 MWRI BRIGHTNESS TEMPERATURE DATA

A technique for estimating tropical cyclone(TC) intensity over the Western North Pacific utilizing FY-3Microwave Imager(MWRI) data is developed. As a first step, we investigated the relationship between the FY-3 MWRI brightness temperature(TB) parameters, which are computed in concentric circles or annuli of different radius in different MWRI frequencies, and the TC maximum wind speed(Vmax) from the TC best track data. We found that the parameters of lower frequency channels' minimum TB, mean TB and ratio of pixels over the threshold TB with a radius of 1.0 or 1.5 degrees from the center give higher correlation. Then by applying principal components analysis(PCA)and multiple regression method, we established an estimation model and evaluated it using independent verification data, with the RMSE being 13 kt. The estimated Vmax is always stronger in the early stages of development, but slightly weaker toward the mature stage, and a reversal of positive and negative bias takes place with a boundary of around 70 kt. For the TC that has a larger error, we found that they are often with less organized and asymmetric cloud pattern, so the classification of TC cloud pattern will help improve the acuracy of the estimated TC intensity, and with the increase of statistical samples the accuracy of the estimated TC intensity will also be improved.

Analysis of morphological characteristics of gravels based on digital image processing technology and self-organizing map

A comprehensive understanding of spatial distribution and clustering patterns of gravels is of great significance for ecological restoration and monitoring.However,traditional methods for studying gravels are low-efficiency and have many errors.This study researched the spatial distribution and cluster characteristics of gravels based on digital image processing technology combined with a self-organizing map(SOM)and multivariate statistical methods in the grassland of northern Tibetan Plateau.Moreover,the correlation of morphological parameters of gravels between different cluster groups and the environmental factors affecting gravel distribution were analyzed.The results showed that the morphological characteristics of gravels in northern region(cluster C)and southern region(cluster B)of the Tibetan Plateau were similar,with a low gravel coverage,small gravel diameter,and elongated shape.These regions were mainly distributed in high mountainous areas with large topographic relief.The central region(cluster A)has high coverage of gravels with a larger diameter,mainly distributed in high-altitude plains with smaller undulation.Principal component analysis(PCA)results showed that the gravel distribution of cluster A may be mainly affected by vegetation,while those in clusters B and C could be mainly affected by topography,climate,and soil.The study confirmed that the combination of digital image processing technology and SOM could effectively analyzed the spatial distribution characteristics of gravels,providing a new mode for gravel research.

A TimeImageNet Sequence Learning for Remaining Useful Life Estimation of Turbofan Engine in Aircraft Systems

Internet of Things systems generate a large amount of sensor data that needs to be analyzed for extracting useful insights on the health status of the machine under consideration.Sensor data of all possible states of a system are used for building machine learning models.These models are further used to predict the possible downtime for proactive action on the system condition.Aircraft engine data from run to failure is used in the current study.The run to failure data includes states like new installation,stable operation,first reported issue,erroneous operation,and final failure.In the present work,the non-linear multivariate sensor data is used to understand the health status and anomalous behavior.The methodology is based on different sampling sizes to obtain optimum results with great accuracy.The time series of each sensor is converted to a 2D image with a specific time window.Converted Images would represent the health of a system in higher-dimensional space.The created images were fed to Convolutional Neural Network,which includes both time variation and space variation of each sensed parameter.Using these created images,a model for estimating the remaining life of the aircraft is developed.Further,the proposed net is also used for predicting the number of engines that would fail in the given time window.The current methodology is useful in avoiding the health index generation for predicting the remaining useful life of the industrial components.Better accuracy in the classification of components is achieved using the TimeImagenet-based approach.

Recent advancement of chemical imaging in pharmaceutical quality control:From final product testing to industrial utilization

Chemical imaging(CI)possesses a strong ability of pharmaceutical analysis.Its great strength relies on the integration of traditional spectroscopy(one dimension)and imaging technique(two dimensions)to generate three-dimensional data hypercubes.Data pre-processing or processing methods are proposed to analyze vast data matrixes and thereby realizing different research objectives.In this review paper,various pharmaceutical applications of quality control over the past few years are summed up in two groups of final product test and industrial utilization.The scope of"quality control"here includes traditional analytical use,process understanding and manufactural control.Finally,two major challenges about undesirable sample geometry and lengthy acquisition time are discussed for prospective commercial or industrial application.

多元纹理图像构造方法及应用

多元图像分析方法在图像展开阶段会丢失像素间空间关联关系,导致其对图像纹理特征的分析能力不足.针对此问题,本文提出了一种基于图像纹理特征的多元图像构造方法,并应用于图像分割.首先,结合滑动窗口法和灰度共生矩阵求取图像各通道的纹理特征影像,叠加纹理特征影像构造多元图像.然后,应用多元图像分析方法对所得多元图像进行分析,分割出感兴趣区域.最后,利用分割结果构造决策树模型,以完成对同类感兴趣区域的分割.在图像数据集上进行仿真实验,实验结果表明,本文所提方法的均交并比(MIoU)与同类方法相比有10%左右的提升.

基于流域日降水量图的相似性搜索方法

为了提升降水量图相似性分析的精确度,提出了一种基于流域日降水量图的相似性搜索方法,该方法从降雨图像中提取日降水量、降雨空间分布和降雨中心特征,并分别计算各特征的相似距离,同时通过提出的归一化折旧累积增益改进粒子群优化的集合加权方法对3个特征的相似距离进行加权融合,作为降雨图像的相似性度量。嘉陵江流域实例验证表明:该方法能够更好地表征降水量图的时空特征,可快速地从降水量图中检索出相似的降雨过程。

冻融胚胎移植中舌象和脉象指标与活产结局的相关性研究

目的探讨冻融胚胎移植(FET)患者舌象和脉象指标与活产结局的相关性,及其与患者内分泌指标的关联性。方法本研究于2021年3月8日-2022年1月5日在上海中医药大学附属曙光医院生殖医学中心进行。根据活产结局将FET患者分为活产组和非活产组。分析两组的内分泌指标[基础促卵泡生成素(b FSH)、基础促黄体生成素(b LH)、基础雌二醇(bE_(2))、基础孕酮(bP)、基础子宫内膜厚度、内膜转换日促卵泡生成素(FSH)、内膜转换日促黄体生成素(LH)、内膜转换日雌二醇(E_(2))、内膜转换日孕酮(P)、内膜转换日子宫内膜厚度]与舌象和脉象指标[舌质(TB)-L,TB-a,TB-b,舌苔(TC)-L,TC-a,TC-b,perAll,perPart,h_(1),h_(4),h_(5),t_(1),h_(1)/t_(1),h_(4)/h_(1)]之间的差异,并用Spearman相关系数分析以上变量之间的相关性。采用多元logistic回归分析方法,筛选基于不同数据集的活产预测模型中的影响因素,包括仅由内分泌指标组成的模型1、仅由舌象和脉象指标组成的模型2、由舌象和脉象指标以及内分泌指标组成的模型3,并对不同数据集模型的有效性进行评价。结果研究纳入活产组78例,非活产组144例。与非活产组相比,活产组的TB-L(P=0.01)和TB-a(P=0.04)升高,b FSH(P=0.01)、perAll(P=0.04)和h_(4)/h_(1) P=0.03]降低。Spearman相关分析显示,活产组的TB-L、TB-b、TC-L、TC-b、perAll、perPart、h_(4)、h_(5)、t_(1)、h_(1)/t_(1)与b FSH、bLH、基础子宫内膜厚度、内膜转换日LH、内膜转换日E_(2)、内膜转换日P、内膜转换日子宫内膜厚度的相关性均有统计学意义(P<0.05)。多元logistic回归分析显示,模型3[曲线下面积(AUC):0.917,95%置信区间(CI):0.863-0.971,P<0.001]优于模型1(AUC:0.698,95%CI:0.593-0.803,P=0.001)和模型2(AUC:0.790,95%CI:0.699-0.880,P<0.001)。将舌象和脉象指标与内分泌指标相结合,建立活产结局的回归方程,最终纳入以下指标:内膜转换日FSH[比值比(OR):0.523,P=0.025]、内膜转换日LH(OR:1.277;P=0.029)、TB-L(OR:2.401,P=0.001)、perPart(OR:1.018,P=0.013)、h_(1)(OR:0.065,P=0.021)、t_(1)(OR:4.354,P=0.024)、h_(4)/h_(1)(OR:0.018,P=0.016)。结论在接受FET治疗的不孕症患者中,舌象和脉象指标与内分泌指标之间存在相关性。舌象和脉象指标的加入显著提高了模型对活产结局的预测能力。具体而言,舌象和脉象指标TB-L、PerPart、h_(1)、t_(1)和h_(4)/h_(1)与最终的活产结局有明显的相关性。

基于多元特征异构集成深度学习的图像识别模型及其应用

随着城市矿产资源循环利用技术的不断发展,废旧手机回收已成为当前研究热点。受限于计算资源和数据资源的相对缺乏,目前基于线下智能回收装备的废旧手机识别精度难以达到实际应用。针对上述问题,提出一种基于多元特征异构集成深度学习的图像识别模型。首先,利用字符级文本检测算法(character region awareness for text detection,CRAFT)提取手机背部字符区域,再利用ImageNet预训练的VGG19模型作为图像特征嵌入模型,利用迁移学习理念提取待回收手机的局部字符特征和全局图像特征;然后,利用局部特征构建神经网络模式光学字符识别(optical character recognition,OCR)模型,利用全局和局部特征构建非神经网络模式深度森林分类(deep forest classification,DFC)模型;最后,将异构OCR和DFC识别模型输出的结果与向量组合后输入Softmax进行集成,基于权重向量得分最大准则获取最终识别结果。基于废旧手机回收装备的真实图像验证了所提方法的有效性。

MRI对局部进展期直肠癌患者术前新辅助放化疗后病理完全反应的预测效能

目的:探讨磁共振成像(magnetic resonance imaging,MRI)对局部进展期直肠癌(locally advanced rectal cancer,LARC)患者术前新辅助放化疗(neo-adjuvant chemoradiotherapy,NCRT)后病理完全反应(pathological complete response,pCR)的预测效能。方法:回顾性纳入2019年11月至2023年10月在陕西省西安市某三甲医院的92例LARC患者,并依据NCRT后手术病理结果分为pCR组(n=21)和非pCR组(n=71)。所有患者NCRT前、后均实施直肠MRI常规扫描及弥散加权成像(diffusion-weighted imaging,DWI)检查,收集其临床特征与影像学特征,运用单因素、多因素Logistic回归分析确定pCR相关影响因素,并基于多因素Logistic回归分析绘制受试者工作特征(receiver operating characteristic,ROC)曲线,运用曲线下面积(area under curve,AUC)分析影像学特征对pCR预测价值。结果:单因素分析结果显示:pCR组NCRT后ADC与△ADC%高于非pCR组,奥沙利铂治疗占比与NCRT前ADC低于非pCR组(P<0.05)。多因素Logistic回归分析结果显示:NCRT前ADC、NCRT后ADC及△ADC%为LARC患者NCRT后pCR的独立影响因素(P<0.05)。ROC曲线分析显示:NCRT前ADC、NCRT后ADC、△ADC%联合预测LARC患者NCRT后pCR的AUC高于单一指标的AUC(P<0.05)。结论:NCRT前ADC、NCRT后ADC及△ADC%为LARC患者NCRT后pCR的独立影响因素。其中NCRT前ADC、NCRT后ADC、△ADC%可作为LARC患者NCRT后pCR的预测因素,且三组参数联合预测价值更为理想。

Subnanometer-resolved chemical imaging via multivariate analysis of tip-enhanced Raman maps

Tip-enhanced Raman spectroscopy(TERS)is a powerful surface analysis technique that can provide subnanometer-resolved images of nanostructures with site-specific chemical fingerprints.However,due to the limitation of weak Raman signals and the resultant difficulty in achieving TERS imaging with good signal-to-noise ratios(SNRs),the conventional single-peak analysis is unsuitable for distinguishing complex molecular architectures at the subnanometer scale.Here we demonstrate that the combination of subnanometer-resolved TERS imaging and advanced multivariate analysis can provide an unbiased panoramic view of the chemical identity and spatial distribution of different molecules on surfaces,yielding high-quality chemical images despite limited SNRs in individual pixel-level spectra.This methodology allows us to exploit the full power of TERS imaging and unambiguously distinguish between adjacent molecules with a resolution of~0.4 nm,as well as to resolve submolecular features and the differences in molecular adsorption configurations.Our results provide a promising methodology that promotes TERS imaging as a routine analytical technique for the analysis of complex nanostructures on surfaces.

A quick and effective multivariate statistical strategy for imaging mass spectrometry

A new multivariate statistical strategy for analyzing large datasets that are produced by imaging mass spectrometry（IMS） techniques is reported.The strategy divides the whole datacube of the sample into several subsets and analyses them one by one to obtain the results.Instead of analyzing the whole datacube at one time,the strategy makes the analysis easier and decreases the computation time greatly.In this report,the IMS data are produced by the air flow-assisted ionization IMS（AFAI-IMS）.The strategy can be used in combination with most multivariate statistical analysis methods.In this paper,the strategy was combined with the principal component analysis（PCA） and partial least square analysis（PLS）.It was proven to be effective by analyzing the handwriting sample.By using the strategy,the m/z corresponding to the specific lipids in rat brain tissue were distinguished successfully.Moreover the analysis time grew linearly instead of exponentially as the size of sample increased.The strategy developed in this study has enormous potential for searching for the mjz of potential biomarkers quickly and effectively.

影响肝细胞癌患者肿瘤微血管侵犯的多因素分析

目的探讨影响肝细胞癌(HCC)患者微血管侵犯(MVI)的危险因素。方法2018年3月~2022年3月我院诊治的HCC患者206例,均接受MRI检查和肝段或肝叶切除术治疗,术后组织病理学检查诊断MVI。应用单因素和多因素Logistic回归分析影响HCC患者MVI形成的独立危险因素。结果在206例HCC患者中,组织病理学检查发现MVI 50例(24.3%);MVI组肿瘤直径≥5 cm和肿瘤细胞低分化分别为56.0%和58.0%,均显著高于非MVI组的32.7%和35.3%(P<0.05),而两组血清AFP水平400 ng/mL、Child-Pugh B级和肿瘤包膜是否完整占比无显著性差异(P>0.05);多因素非条件Logistic回归模型分析果显示,肿瘤≥5 cm和肿瘤细胞呈低分化为HCC患者并发MVI的独立危险因素(OR=1.166,OR=1.141,均P<0.05)。结论肿瘤直径大和细胞低分化为影响HCC患者并发MVI的独立危险因素,术前进行影像学检查了解肿瘤大小或可初步判断HCC患者是否存在MVI而进行必要的治疗准备可能有助于制定合理的手术方案。

基于多元线性回归模型的GF-1 PMS多光谱影像重构方法

目的:多元线性回归模型在保持输入自变量光谱信息和空间特征的同时,通过线性变换获取自变量和因变量的光谱拟合关系,对原输入自变量的光谱信息进行优化,从而获得高空间分辨率和丰富光谱信息的重构数据。方法:利用同期获取的OLI(Operational Land Imager)和PMS(Panchromatic and Multispectral Scanner)多光谱遥感影像,根据最小二乘法构建多元线性回归模型,重构生成具有丰富光谱特征和空间特征的遥感影像,从主客观两个方面评价重构影像的质量。结果:在目视解译(主观)方面,重构影像在一定程度上保留了原OLI影像的光谱特性,提升了原PMS影像的清晰度和分辨性;在量化角度(客观)方面,重构影像的信息量和平均梯度比原OLI对应波段影像的信息量(在部分波段上)和平均梯度要低,但比原PMS影像的信息量和平均梯度要高,可见重构影像的质量介于原PMS影像和OLI影像的质量之间。结论:以青海省门源回族自治县的耕地内不同作物为实例对象,利用最大似然法获取门源县青稞和油菜的空间分布,研究区实测数据验证表明,重构影像对耕地内部青稞与油菜的提取精度高于原PMS和OLI多光谱影像的提取精度。

冬小麦叶绿素含量高光谱检测技术

以大田冬小麦叶绿素含量为研究对象,首先利用高光谱成像系统以线扫描方式获取其反射光谱图像,选择感兴趣区域(ROI)并计算出光谱平均反射率值;然后分别针对其原始光谱和一阶差分光谱,通过相关分析和逐步回归分析,得到能反映叶绿素含量变化的7个最佳优化波长;进而基于该优化波长采用多元线性回归(MLR)方法组建模型,通过假设检验剔除对模型贡献不显著的3个波长变量。选用剩余的4个波长即710.85、767.42、650和520 nm作为自变量重新建立模型,基于校正集和预测集模型的决定系数R2分别为0.843 4和0.709 3。研究结果表明,利用高光谱技术检测大田冬小麦叶绿素含量的方法是可行的。

产品设计中用户感知意象的思维结构

为获取产品总意象、子意象和产品构成元素之间的关系,分析了产品设计中用户感知意象的生成机制,并提出了感知意象思维结构的模型。以太阳镜外形为研究案例,采用多元自适应回归样条技术构建产品总意象与造型、色彩、材质和光影子意象的关系模型,据此应用子意象最低规则和意象值近似规则,在Matlab软件中求解四个子意象的最优组合。采用回归树技术得到各子意象与产品构成元素的关系模型,进一步地以感知意象区间为基础,求解对应于四个最优子意象的最优产品构成元素,据此得到对应于用户感知意象的最优产品构成元素组合。研究表明:在感知意象的思维结构中,产品子意象与构成元素的关系可用树结构表征,而与产品总意象之间存在复杂的映射关系。

基于多变量图像分析的铜矿泡沫浮选分类与识别

针对矿物浮选过程中以人工观测为主的浮选状态监测易受人主观因素影响,长流程的浮选现场难以实时获得生产状态信息,引起在线监测信息的不准确性及滞后,严重影响浮选生产工况及时调整,造成生产过程资源和能源浪费的问题,基于多变量图像分析方法研究矿物加工领域的泡沫浮选过程泡沫图像的分析与特征表征,并融合多变量图像分析、多分辨率分析、多分辨率-多变量图像分析、改进分水岭的图像分割算法和基于模板匹配的宏块跟踪方法,提取铜矿泡沫图像的颜色、纹理、尺寸、速度和稳定度特征.在此基础上,对铜矿泡沫浮选生产状态进行了分类与识别,并建立泡沫图像变量特征与工艺指标之间的关系模型,可用来预测铜矿泡沫品位.应用结果表明:该方法可实现铜矿浮选过程的实时监控及生产状态的识别.

以高光谱数据有效预测苹果可溶性固形物含量

从高光谱数据中选取能够有效进行内部品质检测的特征波长,是利用高光谱成像技术进行水果品质定量分析的关键。本文采用遗传算法(GA)、连续投影算法(SPA)和GA-SPA算法分别从400～1 000nm的苹果高光谱图像中提取特征波长,利用偏最小二乘法(PLS)、最小二乘支撑向量机(LS-SVM)和多元线性回归(MLR)建模进行苹果可溶性固形物含量(SSC)的定量分析并进行了综合比较。160个样品中,120个用于建模,40个用于预测。比较发现SPA-MLR模型获得了最好的结果,R2p,RMSEP和RPD分别为0.950 1,0.308 7和4.476 6。结果表明:SPA能够有效地用于高光谱数据的变量选择,利用SPA-MLR可建立稳健的苹果SSC预测模型,较少的有效变量和MLR模型的易解释性表明该模型在在线检测和便携式仪器开发中具有较大的应用潜力。

基于多元统计模型的分形小波自适应图像去噪

为了克服分形小波去噪方法缺乏保护图像的边缘与细节的缺陷,文中提出了一种基于多元统计模型的分形小波自适应图像去噪算法.通过建立一个参数可以灵活改善的多元统计模型来准确地估计各种相关信息,并通过最小化残差来自适应调整模型参数;在适度的噪声方差下根据拼贴距离找出最好的子树域中的近优父子树,通过使用四叉树分割来实现对噪声图像自适应地预测分形小波无噪图像编码,从而达到去除噪声和保护图像的边缘与细节的目的.实验结果表明,该方法在去除噪声的同时,能有效地保持图像的边缘特征和保留图像的精细结构.由于采用了预测小波分形编码,优化了算法结构,算法的处理速度比较快.因此,完全可以达到实时图像处理过程中的去噪预处理对处理速度的要求.

基于RS流域枯水资源的判读识别

在GIS和RS技术的支持下,以单个水文站所控制的流域为统计单元,借助于ArcView软件强大的计算功能、ERDARSIMAGE软件强大的图象处理功能,在TM影像确定单个水文站所控制的流域面并数字化,通过SPSS软件进行多元回归分析,提出从TM影像判读枯水资源的多元回归方程。