一种陶瓷方形扁平封装外观缺陷检测方法

提出一种基于机器视觉的陶瓷方形扁平封装外观缺陷检测方法。对于封装外形尺寸较大而缺陷较细微的情形,将待检片分为多个区域与标准样片进行比对检测。首先通过Foerstner特征点检测法提取标准片图像的特征点,然后使用随机抽样一致性(RANSAC)图像匹配算法,将所有标准片图像拼接并融合生成一张标准片全幅面模板,再将待检片分区与标准片模板进行序贯比对,以提取可疑区域,最后利用支持向量机(SVM)分类器对可疑区域进行筛选分类。实验结果表明,这种方法不仅克服了传统视觉检测过程中视野范围与图像分辨率相互制约的矛盾,且对陶瓷方形扁平封装表面缺陷具有较高的检出率。     

自适应条件下的相邻区分精度SVM集成模型研究

在数据进行集成的实际过程当中,分类器往往具有自主性,会随着样本数据的变化对自己进行 适 当调整,以此来提高自己的适应能力。对此,本研究以在数据样本区域内对相邻数据进行区分 的方法进行SVM集成方法研究,并最终提出了一种切实可行地支持SVM进行集成的方式。即针 对区分的数据样本区域,以一种新的搜索算法进行研究,利用FCM与模糊贴近度的结合来进行 计 算,实现在模糊特征空间集合频域自身位置的自动确定,再根据各项分类器的阈值数据系统自 行 录用当中的优异数据结果。并最终形成个体分器的数据结果从而进行集合性判定。结果显 示。在减少区分判断用时的前提下,这样一种数据算法能够达到提升分类器功能的有效作用 ;所建立的SVM集成模型具备动态自主适应性。集成过程当中分类器的个数选取关键点在于 分类精度阀值的选取,据此可以通过最优阀值的选取以达到模型判别能力的极大提升。     

随机蕨丛算法下的三点焊缝定位方法

针对在焊接过程中工件易变形、光照干扰等情况下不能精确定位焊缝的问题,提出一种基于方向梯度特征的随机蕨丛算法。首先建立焊缝模型分类器,并用其在测试样本中得到焊缝部分。其次,采用随机蕨丛算法得到所有训练样本的特征类型并计算特征点的平均坐标差值。最后,判断测试样本所属类型,根据坐标偏移量求出特征点具体坐标值来实现定位。实验结果表明：所提出的算法在保证较高精确定位的同时,还具有很好的时效性。     

OW‐SVM: Ontology and whale optimization‐based support vector machine for privacy‐preserved medical data classification in cloud

Cloud is a multitenant architecture that allows the cloud users to share the resources via servers and is used in various applications, including data classification. Data classification is a widely used data mining technique for big data analysis. It helps the learners to discover hidden data patterns by training massive data collected from the real world. Because this trained model is the private asset of an entity, it should be protected from all other noncollaborative entities. Therefore, it is essential to take effective measures to preserve the confidential data. The objective of this paper is to preserve the privacy of the confidential data in the cloud environment by introducing the medical data classification method. In view of that, this paper presents a method for medical data classification using a novel ontology and whale optimization‐based support vector machine (OW‐SVM) approach. Initially, privacy‐preserved data are developed adopting Kronecker product bat approach, and then, ontology is built for the feature selection process. Ontology and whale optimization‐based support vector machine is then proposed by integrating ontology and whale optimization algorithm into SVM, in which ontology and whale optimization algorithm is used for the feasible selection of kernel parameters. The experiment is done using 3 heart disease datasets, such as Cleveland, Switzerland, and Hungarian. In a comparative analysis, the performance of the OW‐SVM approach is compared with that of K‐nearest neighbor, Naive Bayes, decision tree, SVM, and OW‐SVM, using accuracy, sensitivity, specificity, and fitness, as the evaluation metrics. The OW‐SVM approach could achieve maximum performance with accuracy of 83.21%, the sensitivity of 91.49%, specificity of 73%, and fitness of 81.955, outperforming existing comparative techniques.     

Imbalanced SVM‐Based Anomaly Detection Algorithm for Imbalanced Training Datasets

Abnormal samples are usually difficult to obtain in production systems, resulting in imbalanced training sample sets. Namely, the number of positive samples is far less than the number of negative samples. Traditional Support Vector Machine (SVM)‐based anomaly detection algorithms perform poorly for highly imbalanced datasets: the learned classification hyperplane skews toward the positive samples, resulting in a high false‐negative rate. This article proposes a new imbalanced SVM (termed ImSVM)‐based anomaly detection algorithm, which assigns a different weight for each positive support vector in the decision function. ImSVM adjusts the learned classification hyperplane to make the decision function achieve a maximum GMean measure value on the dataset. The above problem is converted into an unconstrained optimization problem to search the optimal weight vector. Experiments are carried out on both Cloud datasets and Knowledge Discovery and Data Mining datasets to evaluate ImSVM. Highly imbalanced training sample sets are constructed. The experimental results show that ImSVM outperforms over‐sampling techniques and several existing imbalanced SVM‐based techniques.     

Research on Link Quality Estimation Mechanism for Wireless Sensor Networks Based on Support Vector Machine

In the application of Wireless sensor net-works (WSNs), effective estimation for link quality is a basic issue in guarantying reliable data transmission and upper network protocol performance. A link quality es-timation mechanism is proposed, which is based on Sup-port vector machine (SVM) with multi-class classification. Under the analysis of the wireless link characteristics, two physical parameters of communication, Receive sig-nal strength indicator (RSSI) and Link quality indicator (LQI), are chosen as estimation parameters. The link qual-ity is divided into five levels according to Packet recep-tion rate (PRR). A link quality estimation model based on SVM with decision tree is established. The model is built on kernel functions of radial basis and polynomial re-spectively, in which RSSI, LQI are the input parameters. The experimental results show that the model is reason-able. Compared with the recent published link quality es-timation models, our model can estimate the current link quality accurately with a relative small number of probe packets, so that it costs less energy consumption than the one caused by sending a large number of probe packets. So this model which is high efficiency and energy saving can prolong the network life.     

面向摔倒监测的纽扣型可穿戴式节点系统

摔倒作为世界第二大意外伤害死亡的诱因,已严重威胁到老年人的身心健康,因而摔倒监测日益变得紧迫。摔倒监测的可穿戴式节点系统的主要目的是监测到摔倒行为并进行及时的报警,从而可以大大减小由摔倒带来的伤害。本系统是基于三轴加速度传感器采集人体的运动信号,利用人体运动时加速度特征的改变,提出了一种准确度高稳定性好的SVM分类算法进行二元分类,区别摔倒与日常活动, 并将结果实时显示在Android手机客户端。实验结果表明,本可穿戴式节点系统能够较好地实现摔倒行为的连续监测,正检率可达94.8%。     

基于加权合成核与三重Markov场的极化SAR图像分类方法

马尔可夫随机场(Markov Random Field,MRF)广泛用于处理遥感图像的分类问题,然而MRF在构建极化合成孔径雷达(Synthetic Aperture Radar,SAR)图像模型时未考虑其非平稳特性且对初始分类较为敏感,为此本文提出了一种基于加权合成核与三重马尔可夫随机场(Triplet Markov Field,TMF)的极化SAR图像分类方法.该方法依据训练样本在特征空间上的距离,提出了加权合成核函数权重系数的自适应确定方法以提高初始分类的精度和普适性;为充分考虑极化SAR图像的非平稳统计特性,利用TMF对极化SAR图像进行统计建模以实现贝叶斯分类.实验结果表明,与基于MRF的极化SAR图像分类方法相比,本文所提方法可获得更高的分类精度和更平滑的同质区域分类结果,而且本文方法能更好地保持图像边缘信息.     

Sparse Penalized Forward Selection for Support Vector Classification

We propose a new binary classification and variable selection technique especially designed for high-dimensional predictors. Among many predictors, typically, only a small fraction of them have significant impact on prediction. In such a situation, more interpretable models with better prediction accuracy can be obtained by variable selection along with classification. By adding an ?₁-type penalty to the loss function, common classification methods such as logistic regression or support vector machines (SVM) can perform variable selection. Existing penalized SVM methods all attempt to jointly solve all the parameters involved in the penalization problem altogether. When data dimension is very high, the joint optimization problem is very complex and involves a lot of memory allocation. In this article, we propose a new penalized forward search technique that can reduce high-dimensional optimization problems to one-dimensional optimization by iterating the selection steps. The new algorithm can be regarded as a forward selection version of the penalized SVM and its variants. The advantage of optimizing in one dimension is that the location of the optimum solution can be obtained with intelligent search by exploiting convexity and a piecewise linear or quadratic structure of the criterion function. In each step, the predictor that is most able to predict the outcome is chosen in the model. The search is then repeatedly used in an iterative fashion until convergence occurs. Comparison of our new classification rule with ?₁-SVM and other common methods show very promising performance, in that the proposed method leads to much leaner models without compromising misclassification rates, particularly for high-dimensional predictors.     

特征级联在车辆检测中的应用与研究

对现有的采用机器学习算法检测车辆进行研究,分析其存在的不足;表现在特征或者算法单一,对光照等条件变化鲁棒性不够;针对这些问题,提出一种融合LBP特征与HOG特征,并结合Adaboost与SVM的车辆检测算法;借鉴级联的思想,首先采用AdaBoost对训练样本提取LBP特征进行训练,得到的分类器用于初步筛选并将其分类结果作为下一层分类器的输入;然后采用SVM算法对训练样本提取HOG特征进行训练,得到的分类器用于二次筛选上一层分类器的分类结果;实验结果证明Adaboost-SVM相结合的办法检测结果精度高,准确率和召回率均达到95%以上,FPPW与FPPI的值均在5%左右;同时由于算法采用的特征对光照条件具有较强的鲁棒性,因此光照条件的变化对算法的识别结果影响较低;实时性方面,每帧图像的处理时间为75 ms,满足实时性要求。