Flame image recognition of alumina rotary kiln by artificial neural network and support vector machine methods

Based on the Fourier transform, a new shape descriptor was proposed to represent the flame image. By employing the shape descriptor as the input, the flame image recognition was studied by the methods of the artificial neural network(ANN) and the support vector machine(SVM) respectively. And the recognition experiments were carried out by using flame image data sampled from an alumina rotary kiln to evaluate their effectiveness. The results show that the two recognition methods can achieve good results, which verify the effectiveness of the shape descriptor. The highest recognition rate is 88.83% for SVM and 87.38% for ANN, which means that the performance of the SVM is better than that of the ANN.

Dynamic Spatial Discrimination Maps of Discriminative Activation between Different Tasks Based on Support Vector Machines

As a set of supervised pattern recognition methods, support vector machines (SVMs) have been successfully applied to functional magnetic resonance imaging (fMRI) field, but few studies have focused on visualizing discriminative regions of whole brain between different cognitive tasks dynamically. This paper presents a SVM-based method for visualizing dynamically discriminative activation of whole-brain voxels between two kinds of tasks without any contrast. Our method provides a series of dynamic spatial discrimination maps (DSDMs), representing the temporal evolution of discriminative brain activation during a duty cycle and describing how the discriminating information changes over the duty cycle. The proposed method was applied to investigate discriminative brain functional activations of whole brain voxels dynamically based on a hand-motor task experiment. A set of DSDMs between left hand movement and right hand movement were reached. Our results demonstrated not only where but also when the discriminative activations of whole brain voxels occurred between left hand movement and right hand movement during one duty cycle.

Research on Application of Support Vector Machine in Machine Learning

In recent years,support vector machine learning methods have gradually become the main research direction of machine learning.The support vector machine has a small structural risk compared with the traditional learning method,which can make the training error and the classifier capacity reach a relatively balanced state.Secondly,it also has the advantages of strong adaptability and strong promotion ability and has been widely praised by the industry.The following discussion focuses on the application of support vector machine in machine learning.

Study on intelligent image welding defects in recognition for the tube pressure vessels

Currently, the welding defects recognition of X-ray nondestructive inspection is principally carried out by manual work, which highly depends on the experience of the inspectors and costs plenty of workload. In this paper, an intelligent image processing and recognition method for the tube welding radiographic testing in large-scale pressure vessels is proposed. Firstly, the raw image is preprocessed by median filtering, pseudo point removing and non-lincar image enhancement. Secondly, the welded joints parts are separated from the whole image by edge detection and threshold segmentation algorithms. Then, the separated images are handled by FFT transformation. Finally, whether defects exist and the specific type of defects are judged by Support Vector Machine. Software developed basing on this method works stably on site, and experiments demonstrate that the recognition results are compliance with the JB/T 4730. 2 or ASME standards.

Diagnosis of sewer pipe defects on image recognition of multi-features and support vector machine in a southern Chinese city

Closed circuit television(CCTV)systems are widely used to inspect sewer pipe conditions.During the diagnosis process,the manual diagnosis of defects is time consuming,labor intensive and error prone.To assist inspectors in diagnosing sewer pipe defects on CCTV inspection images,this paper presents an image recognition algorithm that applies features extraction and machine learning approaches.An algorithm of image recognition techniques,including Hu invariant moment,texture features,lateral Fourier transform and Daubechies(DBn)wavelet transform,was used to describe the features of defects,and support vector machines were used to classify sewer pipe defects.According to the inspection results,seven defects were defined;the diagnostic system was applied to a sewer pipe system in a southern city of China,and 28,760 m of sewer pipes were inspected.The results revealed that the classification accuracies of the different defects ranged from 51.6% to 99.3%.The overall accuracy reached 84.1%.The diagnosing accuracy depended on the number of the training samples,and four fitting curves were applied to fit the data.According to this paper,the logarithmic fitting curve presents the highest coefficient of determination of 0.882,and more than 200 images need to be used for training samples to guarantee the accuracy higher than 85%.

Computer‑aided CT image processing and modeling method for tibia microstructure

We present a method for computed tomography(CT)image processing and modeling for tibia microstructure,achieved by using computer graphics and fractal theory.Given the large-scale image data of tibia species with DICOM standard for clinical applications,we take advantage of algorithms such as image binarization,hot pixel removing and close operation to obtain visually clear image for tibia microstructure.All of these images are based on 20 CT scanning images with 30μm slice thickness and 30μm interval and continuous changes in pores.For each pore,we determine its profile by using an improved algorithm for edge detection.Then,to calculate its three-dimensional fractal dimension,we measure the circumference perimeter and area of the pores of bone microstructure using a line fitting method based on the least squares.Subsequently,we put forward an algorithm for the pore profiles through ellipse fitting.The results show that the pores have significant fractal characteristics because of the good linear correlation between the perimeter and the area parameters in log–log scale coordinates system,and the ratio of the elliptical short axis to the long axis through ellipse fitting tends to 0.6501.Based on support vector machine and structural risk minimization principle,we put forward a mapping database theory of structure parameters among the pores of CT images and fractal dimension,Poisson’s ratios,porosity and equivalent aperture.On this basis,we put forward a new concept for 3D modeling called precision-measuring digital expressing to reconstruct tibia microstructure for human hard tissue.

Research on Feature Fusion Technology of Fruit and Vegetable Image Recognition Based on SVM

In order to improve the accuracy and stability of fruit and vegetable image recognition by single feature, this project proposed multi-feature fusion algorithms and SVM classification algorithms. This project not only introduces the Reproducing Kernel Hilbert space to improve the multi-feature compatibility and improve multi-feature fusion algorithm, but also introduces TPS transformation model in SVM classifier to improve the classification accuracy, real-time and robustness of integration feature. By using multi-feature fusion algorithms and SVM classification algorithms, experimental results show that we can recognize the common fruit and vegetable images efficiently and accurately.

基于CT图像特征提取及改进SVM算法的肺结节诊断方法研究

目的:提出一种基于CT图像特征提取及改进支持向量机(support vector machine,SVM)算法的肺结节诊断方法,以提高肺结节自动识别的准确性和效率。方法:采用融合深度学习特征提取与传统手工提取的级联特征提取方法进行CT图像特征提取,并将提取特征输入改进SVM算法中完成肺结节自动化识别,使用多核学习支持向量机(multiple kernel learning support vector machine,MKL-SVM)算法与融合模拟退火(simulated annealing,SA)算法改进的粒子群(particle swarm optimization,PSO)算法(PSO-SA算法)进行参数寻优。通过传统特征提取、深度学习特征提取与级联特征提取进行对比,测试级联特征的性能。使用单核函数(RBF核、Sigmoid核以及多项式核函数)进行对比测试,以验证MKL-SVM算法的性能。为验证PSO-SA算法的有效性,使用Sigmoid核的SVM函数进行测试,比较PSO算法和PSO-SA算法寻优的适应度曲线。在同一数据集下与现有肺部计算机辅助诊断(computer aided diagnosis,CAD)模型进行对比,以验证提出的级联特征结合改进MKL-SVM(cascade features+improved MKL-SVM,CF+MKL-SVM)模型的诊断效能。结果:级联特征性能测试结果表明,级联特征提取的F值的平均值为0.9341、最大值为0.9573、最小值为0.9195、中位数为0.9397,均高于手工特征提取和深度学习特征提取的精准度。核函数对比测试结果表明,MKL-SVM算法F值的平均值为0.9243、最大值为0.9350,AUC为0.9873,诊断效能最优。Sigmoid核对比测试结果表明,PSO-SA算法最佳适应度值为0.9437,相较于PSO算法有更优的性能表现。模型对比测试结果表明,与肺部CAD模型相比,CF+MKL-SVM模型具有很高的泛化能力,AUC为0.9845,且各项指标均保持在0.9以上,特异度和精准度均较高。结论:提出的方法能够对肺癌进行自动化识别并提高了肺癌检测的准确率。

Hepatic CT image retrieval based on the combination of Gabor filters and support vector machine

Content-based image retrieval has been an active area of research for more than ten years. Gabor schemes and support vector machine （SVM） method have been proven effective in image representation and clas-sification. In this paper, we propose a retrieval scheme based on Gabor filters and SVMs for hepatic computed tomography （CT） images query. In our experiments, a batch of hepatic CT images containing several types of CT findings are used for the retrieval test. Precision comparison between our scheme and existing methods is presented.

胸部CT图像中孤立性肺结节良恶性快速分类

为突破医学影像诊断学依据医学征象进行定性诊断准确度不高的瓶颈,针对胸部CT图像中孤立性肺结节(SPN)定性诊断问题,提出了能够用图像特征有效表示SPN病理特性,快速准确诊断SPN良恶性的计算机辅助诊断系统。采取交互式分割方法从胸部CT图像中提取出SPN;直接计算SPN图像的多分辨率直方图得到768维空间信息特征样本集;然后,充分利用具有处理高维数据集优势的支持向量机(SVM)构造SPN良恶性分类器;最后,通过测试样本集对经训练后的SVM分类器进行测试以评价分类性能。对214例病例进行实验,结果表明:240个SPN图像的768维特征计算所用时间为4.83s,SVM分类器训练测试所用时间为2.24s,敏感性为73.33%,特异性为70%,准确度达71.67%,接受者操作特性曲线(ROC)下面积(AUC)为0.7864。该系统提取的高维图像空间信息特征能够有效表示SPN特性;没有考虑医学征象进行SPN定性诊断的准确度即可达到71.67%,同时分类速度比传统纹理算法提高了近50倍,为医学影像学解决SPN定性诊断问题提供了便捷、客观的辅助手段。

基于CT扫描图像的碳酸盐岩油藏孔隙分类方法

碳酸盐岩油藏具有复杂的储集空间和油气渗流特征,定量描述油藏中孔隙、裂缝、孔洞等储集空间的大小、形状及连通性难度较大。提出了一种基于扫描图像判断碳酸盐岩孔隙类型方法,可定量表征孔隙参数,并对岩心样品进行自动分类。该方法首先对碳酸盐岩的岩心扫描图像进行灰度转换和提高信噪比的预处理,然后对图像进行分割,区分出孔隙区域与基质区域。在此基础上,通过形态学处理和特征参数计算等步骤提取出孔隙特征参数,根据特征参数建立特征向量,采用支持向量机方法对CT图像中的孔隙、孔洞和裂缝进行自动识别并分类。在对岩心所有截面孔隙识别的基础上,提出了判断岩心孔隙类型的分类指数。T油田M油藏和F油藏应用结果表明:该方法识别精度较高,有效确定了油藏中占主导地位的孔隙类型,对油田有效开发具有一定的指导意义。

支持向量机与区域增长相结合的CT图像并行分割

针对经典区域增长算法中生长规则确定的困难和单纯使用支持向量机分割速度慢的问题,提出了一种支持向量机与区域增长相结合的图像并行分割方法。首先,从已知分割结果的图像中选取一定数量的目标区域与非目标区域样本点作为支持向量机分类器的训练样本并训练支持向量机,然后利用训练好的支持向量机自动寻找种子点并进行区域增长,在区域增长过程中使用支持向量机分类器作为增长规则,最后,针对边缘和噪声像素点进行必要的后处理。测试实验获得了较好的分割效果和较快的分割速度且能实现自动分割,表明所提出的方法是可行有效的。

CT图像中肿大淋巴结肺癌转移分类方法

为解决肺癌N分期中胸部CT难于对肿大淋巴结是否癌转移进行评价的问题,寻求能够有效表示淋巴结病理特性的图像特征,实现对肿大淋巴结癌转移快速准确地判别。该文采取交互式分割从CT图像中提取出肿大淋巴结;直接计算淋巴结的多分辨率直方图得到200维空间信息特征样本集;利用具有处理高维数据集优势的支持向量机(SVM)构造分类器;用测试集对经训练的SVM分类器进行测试以评价分类性能。经96例病例实验结果表明:100个淋巴结图像的200维特征计算用时1.91s,SVM分类器训练测试用时1.36s,敏感性76%,特异性64%,准确度70%,接受者操作特性曲线(ROC)下面积(AUC)0.6525。高维图像空间信息特征能够有效表示淋巴结特性;没有考虑医学征象进行肿大淋巴结癌转移定性诊断的准确度就达到了70%,同时分类速度比传统纹理算法提高了约10倍。

基于Hessian矩阵和支持向量机的CT图像裂纹分割

工业CT裂纹分割是工业CT图像处理中的一项关键技术,而CT图像中的伪影、噪声等干扰会对裂纹分割带来极大的困扰。为了提高CT图像中裂纹的分割精度,本文通过分析CT图像中裂纹的特征,提出了结合Hessian矩阵和支持向量机的CT图像裂纹识别与分割的方法。首先采用基于Hessian矩阵的多尺度滤波方法提取CT图像中的线状结构并对提取的线状结构进行对比度增强;然后利用图像像素点灰度在空间的分布规律提取线状结构图像的纹理特征信息;再使用基于径向基函数的支持向量机训练裂纹识别分类器;进而使用训练好的模型定位测试图像中裂纹所在的方块区域;最后使用自动阈值分割算法得到CT图像中的裂纹。实验结果表明,结合Hessian矩阵和支持向量机的分割方法能够抑制图像中非目标区域,提高了算法的抗干扰性,对裂纹子图像识别的准确率可达94.5%,具有实际的工程应用价值。

基于LDA和SVM的肺结节CT图像自动检测与诊断

在已实现疑似肺结节图像分割的基础上,提取肺结节图像多维特征,应用LDA和SVM统计分类器,通过对大量样本的训练,实现对肺结节CT图像的自动检测和诊断。实验结果表明该检测和诊断方法达到了较好的分类效果,是适用于CAD系统的实用方法。

基于CT三维图像的肺结节良恶性鉴别研究

目的运用计算机方法处理肺部CT图像以识别肺结节良恶性并辅助肺癌诊断,现已成为国内外研究的热点。方法提出一种基于肺部CT图像三维肺结节信息的肺结节良恶性鉴别方法。首先结合阈值分割、区域生长、形态学运算等在CT图像上分割出肺结节,进而提取每个肺结节的三维特征并优化,选择有效特征。然后,基于有效特征采用支持向量机(support vector machine,SVM)的分类算法对多维向量所描述的肺结节进行良恶性的二分类。最后从敏感性、特异性、准确率以及似然比等方面全面评估分类结果。结果实验获得敏感性为0.7776,准确性为0.7378,阳性似然比2.2410,阴性似然比0.3682,显示基于CT三维肺结节图像可以达到令人满意的肺部肿瘤良、恶性鉴别效果。结论上述结果证明了基于CT三维图像的肺结节良恶性鉴别方法的可行性。本研究对计算机辅助肺癌的诊断具有重要意义。

基于改进LeNet-5网络的堆芯燃料组件编码识别

在核电站堆芯核燃料组件水下组装作业中,需要通过视觉技术进行组件编码的识别以便准确定位组件的安装位置。针对水下环境中弱光照等问题导致了图像质量的降低,本文通过乘方增强算法、OSTU算法、CLAHE算法和拉普拉斯变换的方法来实现堆芯燃料组件编码字符水下图像的增强。为了提高编码识别效果,提出了一种整合LeNet-5网络和支持向量机(SVM)的模型,在网络中添加BN(Batch Normalization)层与Dropout层来加速网络的运行速度,并改进Sigmoid函数,增加函数的平滑性,以此来减少梯度消失。实验表明,在自定义数据集上的验证准确率为99.82%,识别率为100%,相比于其他模型有显著的提升。

基于支持向量机的肝纤维化CT图像分类

基于统计学习理论中结构风险最小化原则的支持向量机是易于小样本的机器学习方法。本文使用支持向量机和二叉树的方法对肝纤维化CT图像进行分类,并与k近邻法和BP神经网络等其它算法进行比较,结果显示对于肝纤维化图像,支持向量机的分类效果和鲁棒性要高于其他两种算法。

基于CT图像的肺结节检测与识别

目的:将肺结节从含有背景、噪声的胸腔区域里检测并识别出来。方法:首先,将DICOM格式的医学图像转换成JPG图像后,应用区域生长法分割出肺实质区域,去掉肺区外的干扰信息。然后,利用多尺度高斯滤波器增强图像后,应用模糊C均值聚类算法提取肺结节感兴趣区域。最后,对肺结节特征进行提取及归一化处理,应用支持向量机分类器识别并标记出肺结节。结果:在随机抽取的120例图像中,检测肺结节的准确率达到92.3%,分类识别肺结节的准确率达到95.6%。实验结果表明,本文方法有效地排除了交叉状和条形状血管等干扰,实现了肺结节的精确检测和识别。结论:本方法在保证检测和识别出正确结节的前提下,降低了误判率,算法也得到了较好的收敛。

基于支持向量机的并行CT图像分割方法

提出了一种基于区域增长和支持向量机的自动并行CT图像分割方法.传统的种子生长方法速度较快,但难于自动获得种子点;而单纯的支持向量机分割准确,但速度较慢.为了解决上述问题,本文将两种方法相结合:首先,训练支持向量机用于分类;然后用支持向量机判断种子点并使用曲率流滤波器进行降噪以光滑图像边缘;最后使用阈值区域生长进行分割.在基于Torque的并行环境下进行的实验证明了本方法的分割效果和速度都优于传统方法.