结合形态学属性断面与支持向量机的合成孔径雷达图像变化检测

针对传统合成孔径雷达(SAR)图像变化检测方法存在误差大、检测率低等问题,提出了一种基于形态学属性断面(MAP)的SAR图像变化检测方法。该方法利用MAP算法提取差异图像的几何结构特征,构造深入描述图像结构化信息的特征向量空间;在利用阈值法对图像进行分割的基础上,引入偏移因子,实现训练样本的自动选取;最后,用支持向量机(SVM)在多维特征空间中对图像进行变化类与非变化类的分类。实验结果显示:本文算法的检测结果优于基于高斯模型的KI阈值法(GM_KI)、基于广义高斯模型的KI阈值法(GGM_KI)和大津法(Otsu)等3种阈值法的检测结果,Kappa系数保持在0.87以上;当峰值信噪比(PSNR)介于[29,44]dB时,抗噪性能指标保持在0.97以上。这些结果证明了文中方法的有效性和优越性。     

基于预分割轮廓的CT图像亚体素表面检测方法

针对基于CT(computed tomography)图像检测分析中的点云提取精度与完整性问题,提出一种基于预分割轮廓的高精度、高完整性的亚体素表面检测方法。首先采用Otsu分割算法提取CT图像的体素级轮廓点集,并以此作为粗定位轮廓自适应地生成用于亚体素表面检测的完备感兴趣区域(region of interest,ROI);然后提出一种基于梯度非极大值抑制的表面体素判定方法,避免了梯度阈值选择难题;最后基于3D Facet模型定位亚体素级表面点位置。实验结果表明,该方法能有效改善传统亚体素检测方法的轮廓丢失、伪边严重等问题,轮廓定位误差小于0.2个体素,同时能够取得3倍以上的计算加速比。     

Microscopic retinal blood vessels detection and segmentation using support vector machine and K-nearest neighbors

The retina is the deepest layer of texture covering the rear of the eye, recorded by fundus images. Vessel detection and segmentation are useful in disease diagnosis. The retina's blood vessels could help diagnose maladies such as glaucoma, diabetic retinopathy, and blood pressure. A mix of supervised and unsupervised strategies exists for the detection and segmentation of blood vessels images. The tree structure of retinal blood vessels, their random area, and different thickness have caused vessel detection difficulties at machine learning calculations. Since the green band of retinal images conveys more information about the vessels, they are utilized for microscopic vessels detection. The current research proposes an administered calculation for segmentation of retinal vessels, where two upgrading stages depending on filtering and comparative histogram were applied after pre-processing and image quality improvement. At that point, statistical features of vessel tracking, maximum curvature and curvelet coefficient are extracted for each pixel. The extracted features are classified by support vector machine and the k-nearest neighbors. The morphological operators then enhance the classified image at the final stage to segment with higher accuracy. The dice coefficient is utilized for the evaluation of the proposed method. The proposed approach is concluded to be better than different strategies with a normal of 92%.     

Automatic measurement of volume percentage stroma in endometrial images using texture segmentation

The popularity of digital microscopy and tissue microarrays allow the use of high-throughput imaging for pathology research. To coordinate with this new technique, it is essential to automate the process of extracting information from such high amount of images. In this paper, we present a new model called the Subspace Mumford-Shah model for texture segmentation of microscopic endometrial images. The model incorporates subspace clustering techniques into a Mumford-Shah model to solve texture segmentation problems. The method first uses a supervised procedure to determine several optimal subspaces. These subspaces are then embedded into a Mumford-Shah objective function so that each segment of the optimal partition is homogeneous in its own subspace. The method outperforms a widely used method in bioimaging community called k-means segmentation since it can separate textures which are less separated in the full feature space, which confirm the usefulness of subspace clustering in texture segmentation. Experimental results also show that the proposed method is well performed on diagnosing premalignant endometrial disease and is very practical for segmenting image set sharing similar properties.     

Automated lung nodule detection and classification based on multiple classifiers voting

Lung cancer is the most common cause of cancer‐related death globally. Currently, lung nodule detection and classification are performed by radiologist‐assisted computer‐aided diagnosis systems. However, emerged artificially intelligent techniques such as neural network, support vector machine, and HMM have improved the detection and classification process of cancer in any part of the human body. Such automated methods and their possible combinations could be used to assist radiologists at early detection of lung nodules that could reduce treatment cost, death rate. Literature reveals that classification based on voting of classifiers exhibited better performance in the detection and classification process. Accordingly, this article presents an automated approach for lung nodule detection and classification that consists of multiple steps including lesion enhancement, segmentation, and features extraction from each candidate's lesion. Moreover, multiple classifiers logistic regression, multilayer perceptron, and voted perceptron are tested for the lung nodule classification using k‐fold cross‐validation process. The proposed approach is evaluated on the publically available Lung Image Database Consortium benchmark data set. Based on the performance evaluation, it is observed that the proposed method performed better in the stateof the art and achieved an overall accuracy rate of 100%.     

基于神经网络边缘提取的工业断层成像图像拟合

通过工业计算机断层成像图像边缘拟合,可获得矢量化的曲线,进而获得工件的计算机辅助设计图纸,实现基于工业断层成像的机械零部件逆向设计。在用两组细胞神经网络对图像进行分割的基础上,进行边缘跟踪和曲线的多维拟合。通过横截圆和轴线的拟合,实现对圆柱形目标的拟合。轴线拟合时,将各层圆心坐标分别投影到xz平面和yz平面进行最小二乘拟合,以降低计算复杂度。对发动机切片图像进行实验,根据拟合得到的参数得出了圆柱形目标的计算机辅助设计图,其拟合均方误差小于0.3像素2。对不规则目标,讨论了基于最小二乘法的分段三次曲线拟合方法在边缘曲线拟合中的应用;对发动机切片图像目标区域进行实验的拟合均方误差均小于0.6像素2。实验结果和误差分析证明,文中的拟合方法是有效的,实现了基于工业断层成像的逆向设计所必需的位图矢量化。     

Accelerometer-based fall detection using feature extraction and support vector machine algorithms

Falls by the elderly may result in hip fractures, paraplegia, and even death. Hence, over the past few decades, considerable research has been conducted on fall detection. Here, an accelerometer-based fall detector is reported that is fastened to a person's waist and includes an accelerometer, a multiplexer, a fifth-order low-pass Butterworth filter, and a microcontroller. Acceleration sensing, noise filtering, and analog-to-digital conversion were performed by the circuitry. The processed signal was sent to a personal computer through Bluetooth and analyzed by customized software. The fall detection algorithm included feature extraction and a support vector machine algorithm for classifying the features. Twenty volunteers performed 12 trials of 6 daily activities and 6 fall events. The results show that the algorithm had high sensitivity (95%) and specificity (96.7%). Thus, this device is expected to have significant application for fall detection.     

Bearing running state recognition based on non-extensive wavelet feature scale entropy and support vector machine

In order to effectively recognize the bearing running state, a new method based on non-extensive wavelet feature scale entropy and the Morlet wavelet kernel support vector machine (MWSVM) was proposed. Firstly, the gathered vibration signals were decomposed by the wavelet to obtain the corresponding wavelet coefficients. Then, based on the integration of non-extensive entropy and the coefficients, the features were extracted by the wavelet feature scale entropy. However, the extracted features remained high-dimensional and excessive redundant information still existed. Therefore, the manifold learning algorithm locality preserving projection (LPP) was introduced to extract the characteristic features and to reduce the dimension. The extracted characteristic features were inputted into the MWSVM to train and construct the running state identification model; the bearing running state identification was thereby realized. Cases of test and actual fault were analyzed. The results validate the effectiveness of the proposed algorithm.     

Hydrodynamics analysis in micro-channels of a viscous coupling using gamma-ray computed tomography

In this work, high-resolution gamma-ray computed tomography (HireCT) was applied for the first time on a viscous coupling to visualize the internal working fluid distribution. The HireCT measurement system comprises a ¹³⁷Cs isotopic source and a gamma-ray detector arc operated in single photon counting mode. It is able to produce cross-sectional images of dense objects with a spatial resolution of about 2 mm. To scan fast rotating parts time-averaged rotation-synchronized CT scanning mode was employed in these experiments. The analyzed viscous coupling (Visco^® clutch of MAHLE Behr) mainly consists of a driven primary disc and a secondary housing with an engine cooling fan mounted on it. The viscous coupling is assembled within an experimental rig. Its primary and secondary parts are axially assembled and a working liquid is pumped into engaged radial ring profiles to provide a defined torque transfer. The internal ring channel width, where the working liquid is to be observed, is considerably lower than one millimeter. Although the HireCT measurement system is not able to resolve these micro-channels, the working liquid can be successfully visualized via its phase contrast. Investigations have been performed at different filling levels corresponding to different transmission slips of the test viscous coupling. Moreover, both radial and tangential liquid distributions for different operational steady states could be determined. Finally, the obtained experimental data were compared to results from computational fluid dynamics (CFD) simulations in some operating points and were in good agreement.     

基于最小二乘支持向量机和机电综合特征的发电机故障诊断

提出了1种最小二乘支持向量机和机电综合特征相结合的发电机故障诊断模型.用二次损失函数代替传统支持向量机中的不敏感损失函数,将不等式约束条件变为等式约束,从而将二次规划问题转变为线性方程组求解,降低了计算的复杂性.提取发电机故障中的综合特征,即振动信号和电流信号,整理后作为诊断模型的特征值,从而得到了故障的典型特征,提高了诊断的准确率.最后从SDF 9型模拟发电机中实测数据进行分析,结果表明,与常规的方法相比,该模型具有较高的分类速度和较好的故障诊断准确率.     

Porosity measurements by X-ray computed tomography: Accuracy evaluation using a calibrated object

Accurate identification and measurement of internal voids and porosity is an important step towards improvement of production processes to obtain high quality materials and products. Recently, the importance of knowing the exact size, shape, volume and location of defects has become even higher as tighter requirements and new standards have been introduced in industry. There are several well-established methods for defects evaluation based on various principles (both destructive and non-destructive). However, all conventional methods have various deficiencies and the information about internal voids/porosity that can be extracted is limited. Most of these drawbacks can be overcome by using X-ray computed tomography (CT). Unlike other methods, CT provides full three-dimensional information about shape, size and distribution of internal voids and porosity; however, the accuracy of measurements is still under investigation. Hence, further evaluations on CT porosity measurements must be performed in order to consider X-ray computed tomography a reliable instrument for the assessment and detection of internal defects.A reference object with artificial defects was used in this research work in order to evaluate the accuracy of porosity measurements by CT. The reference object was manufactured by ultra-precision micro-milling. The object contains dismountable components with embedded internal hemispherical features that simulate internal porosity. The artificial porosity was micro-milled on top surfaces of dismountable cylindrical inserts. The hemispherical calottes were thereafter calibrated by traceable coordinate measuring systems and calibrated values were compared to actual values measured by a CT system. The accuracy of CT porosity measurements was then evaluated based on results obtained on various measurands, using different software tools and measuring procedures, comparing real scans to numerical simulations and investigating the influence of CT system parameters settings on measurement results.     

Rolling element bearing fault detection using support vector machine with improved ant colony optimization

In support vector machine (SVM), it is quite necessary to optimize the parameters which are the key factors impacting the classification performance. Improved ant colony optimization (IACO) algorithm is proposed to determine the parameters, and then the IACO-SVM algorithm is applied on the rolling element bearing fault detection. Both the optimal and the worst solutions found by the ants are allowed to update the pheromone trail density, and the mesh is applied in the ACO to adjust the range of optimized parameters. The experimental data of rolling bearing vibration signal is used to illustrate the performance of IACO-SVM algorithm by comparing with the parameters in SVM optimized by genetic algorithm (GA), cross-validation and standard ACO methods. The experimental results show that the proposed algorithm of IACO-SVM can give higher recognition accuracy.     

基于最小二乘支持向量机的绝缘子盐密光纤在线检测

本文介绍了利用光纤技术在线检测绝缘子盐密的基本原理,并在对受污石英玻璃棒和绝缘子进行试验研究的基础上,针对光通量衰减与盐密、环境湿度和尘埃比率之间复杂的非线性关系,利用最小二乘支持向量机,建立了以光通量衰减、环境湿度、尘埃比率等3个变量作为输入参数,等值附盐密度作为输出参数的智能预测模型,并通过部分实验数据验证了模型的有效性。最后应用预测模型开发了绝缘子盐密在线检测系统。     

Iodine potassium iodide improves the contrast‐to‐noise ratio of micro‐computed tomography images of the human middle ear

High‐resolution imaging of middle‐ear geometry is necessary for finite‐element modeling. Although micro‐computed tomography (microCT) is widely used because of its ability to image bony structures of the middle ear, it is difficult to visualize soft tissues – including the tympanic membrane and the suspensory ligaments/tendons – because of lack of contrast. The objective of this research is to quantitatively evaluate the efficacy of iodine potassium iodide (IKI) solution as a contrast agent. Six human temporal bones were used in this experiment, which were obtained in right‐left pairs, from three cadaveric heads. All bones were fixed using formaldehyde. Three bones (one from each pair) were stained in IKI solution for 2 days, whereas the other three were not stained. Samples were scanned using a microCT system at a resolution of 20 μm. Eight soft tissues in the middle ear were segmented: anterior mallear ligament, incudomallear joint, lateral mallear ligament, posterior incudal ligament, stapedial annular ligament, stapedius muscle, tympanic membrane and tensor tympani muscle. Contrast‐to‐noise ratios (CNRs) of each soft tissue were calculated for each temporal bone. Combined CNRs of the soft tissues in unstained samples were 6.1 ± 3.0, whereas they were 8.1 ± 2.7 in stained samples. Results from Welch's t‐test indicate significant difference between the two groups at a 95% confidence interval. Results for paired t‐tests for each of the individual soft tissues also indicated significant improvement of contrast in all tissues after staining. Relatively large soft tissues in the middle ear such as the tympanic membrane and the tensor tympani muscle were impacted by staining more than smaller tissues such as the stapedial annular ligament. The increase in contrast with IKI solution confirms its potential application in automatic segmentation of the middle‐ear soft tissues.     

Areal surface texture data extraction from X-ray computed tomography reconstructions of metal additively manufactured parts

Many applications that exploit the manufacturing flexibility of additive manufacturing (AM) produce surfaces, primarily internal features, which cannot be measured using conventional contact or line-of-sight optical methods. This paper evaluates the capability of a novel technique to extract areal surface data from micro-focus X-ray computed tomography (XCT) from AM components and then generate surface parameter data per ISO 25178-2. This non-destructive evaluation of internal features has potential advantages during AM product research and commercial production. The data extracted from XCT is compared with data extracted using a focus variation instrument. A reference dimensional artefact is included in all XCT measurements to evaluate XCT surface determination performance and dimensional scaling accuracy. Selected areal parameters generated using the extraction technique are compared, including Sa, for which the nominal difference between the value obtained using XCT and used the focus variation method was less than 2.5%.     

On-machine tool prediction of flank wear from machined surface images using texture analyses and support vector regression

In this paper, a method for on-machine tool condition monitoring by processing the turned surface images has been proposed. Progressive monitoring of cutting tool condition is inevitable to maintain product quality. Thus, image texture analyses using gray level co-occurrence matrix, Voronoi tessellation and discrete wavelet transform based methods have been applied on turned surface images for extracting eight useful features to describe progressive tool flank wear. Prediction of cutting tool flank wear has also been performed using these eight features as predictors by utilizing linear support vector machine based regression technique with a maximum 4.9% prediction error.