分布式网络下改进的Top-k查询算法

现有Top-k查询算法主要运用在集中式关系型数据库中,当应用于分布式网络时会产生巨大的通信开销,导致算法效率低下。为此,提出一种改进的Top-k查询算法,利用预处理索引表对分布式网络中无关数据进行裁剪,在此基础上建立包含正确Top-k结果的候选子集并实现Top-k查询。实验结果表明,与Fagin和Naive Top-k查询算法相比,改进算法获得的查询结果更准确,运行时间更短,网络开销更小。     

基于虚拟仪器的扁钢内部缺陷超声三维成像方法研究

超声三维成像技术因其能够提供人体结构的立体信息,而被广泛的应用于医学诊断领域;然而,由于计算量大、成本高、速度慢,工业构件的超声三维成像技术却鲜见报道;弹簧扁钢是国民经济建设的重要钢材品种之一,通过超声三维成像技术测量扁钢内部缺陷的三维分布对于提高扁钢质量控制能力具有重要作用;为提高三维成像效率及降低成像系统的研发周期,提出高效的数据采集及三维图像重构方法,并基于虚拟仪器技术开发了三维图像重构软件;研究结果显示,基于超声水浸聚焦分层C扫描的数据提取方法能够满足三维成像的要求,所提出的数据重构方法及基于虚拟仪器的三维成像软件能够准确重构扁钢内部层片状缺陷的三维分布,是弹簧扁钢内部缺陷评价的有效方法。     

基于注意力机制的三维点云车辆目标检测

针对自动驾驶场景下三维点云车辆的识别和定位问题,提出了一种基于注意力机制的三维点云车辆目标检测算法.算法将稀疏无序的点云空间划分成等距规则的体素表示,用三维稀疏卷积和辅助网络同步从所有体素中提取内部点云特征,进而生成鸟瞰图.但在将内部三维的点云特征转化为二维的鸟瞰图后,通常会造成目标空间特征信息丢失,使得最终检测结果以及方向性预估差.为进一步提取鸟瞰图中特征信息,提出了一种注意力机制模块,其中包含两种注意力模型,并对其采用首、中、尾的"立体式"布局结构,实现对鸟瞰图中特征信息的放大和抑制,最后使用卷积神经网络和PS-Warp变换机制对处理过后的鸟瞰图进行三维目标检测.实验表明,该算法在保证实时检测效率的前提下,与现有算法相比,具有更好的方向预估性以及更高的检测精度.     

木材彩色图像缺陷分割——基于Gabor滤波的改进C-V彩色模型

分析了木材节子缺陷、单板节子的特点,提出了一种基于多通道Gabor滤波的改进C-V彩色模型的木材缺陷识别算法。该算法将彩色图像作为一个整体的图像,保留了图像的彩色信息。该算法利用多通道Gabor滤波器、K-均值聚类算法得到缺陷目标与背景的彩色区分图像;利用改进的彩色C-V模型对新图像进行边缘提取,得到理想的实验结果。与采用基于改进C-V模型与小波变换的灰度图像缺陷识别算法相对比,结果表明该方法可快速、准确地实现对木材节子缺陷彩色图像及单板多节子彩色图像的分割。     

基于OpenGL的多臂井径仪三维成像系统的设计

文章介绍了多臂井径仪的测量原理,设计了基于OpenGL的多臂井径仪三维成像系统,详细介绍了多臂井径仪三维成像系统空间点的获得、数据的处理及软件的设计、三维重构的实现方法。从三维重构的整体效果图可以看出,该系统能准确、快捷地实现套管的三维重构,算法简捷合理。     

地震资料逆时偏移中的图形处理器加速算法

叠前逆时偏移（RTM）方法是目前地震勘探领域最为精确的一种地震数据成像方法,其运用双程声波方程进行波场延拓,可实现对复杂构造介质的准确成像．文中采用互相关成像条件对震源波场与检波点波场在同时刻相关成像．针对RTM方法计算量大的问题,将图形处理器（GPU）引入到RTM计算中,充分挖掘GPU的众核结构优势,利用基于CUDA架构的并行加速算法取代传统CPU的串行运算,对逆时偏移算法中较为耗时的波场延拓和相关成像过程进行加速．复杂模型测试结果表明,在确保RTM成像精度的前提下,相比于传统CPU计算,GPU并行加速算法可大幅度地提高计算效率,进而实现基于GPU加速的叠前逆时偏移算法对复杂介质的高效率、高精度成像．     

应用于Lamb波的最小功率无畸变波束形成算法

在利用Lamb波进行板状结构损伤检测时,采用经典最小功率无畸变波束形成算法可以获得较高的损伤成像精度,但仍无法避免频散效应对成像精度的影响,且该算法存在谱矩阵求逆运算复杂、奇异谱矩阵无法求逆等缺陷,显著降低计算效率。为解决上述问题,本文提出了更适用于Lamb波检测的改进型最小功率无畸变波束形成算法。该算法在频域进行波束形成以解决频散对成像的影响,并结合最小二乘递推法与对角加载法来进行谱矩阵求逆,以提高计算效率。实验及仿真结果表明,该算法能有效去除频散对损伤成像结果的影响,从而有效地解决了传统波束形成算法的成像结果中存在伪影,且成像分辨率低的问题;同时解决了谱矩阵求逆困难的问题,使得计算时间显著缩减。     

基于Skyline服务的Top-k选择方法

为缩小Skyline服务集,提高服务选择的效率,提出一种Skyline服务Top-k选择方法。首先,用数据推理的方式为Skyline服务的Top-k选择提供理论依据,并提出Skyline服务Top-k选择的相关命题;然后,基于这些命题,提出Skyline服务Top-k选择算法,该算法可以得到被选择可能性最大的Top-k Skyline服务集;最后,通过实验证明,该方法能有效降低服务选择的时间,而不影响服务组合的最终结果。     

基于空气耦合超声Lamb波对铝板的特征成像检测

A0模态Lamb波因具有低衰减、短波长、对缺陷敏感等优良特性被用于检测大面积层板结构。为了实现铝板中缺陷的检测与重构,提出了一种基于空气耦合Lamb波扫描成像检测技术,对存在冲击损伤的铝板进行扫描检测。扫描时,激励与接收传感器置于铝板同侧并倾斜布置,沿着两个相互正交的方向进行Lamb波扫查,对正交路径下的检测信号进行快速傅立叶变换(FFT),将得到的频域信号与基准频域信号之间的信号差异系数（SDC）作为特征值,并结合损伤概率成像算法和数据融合算法进行缺陷的重构和成像。实验结果表明,基于损伤指数的特征值成像方法可以有效表征铝板缺陷,成像结果与缺陷实际位置非常吻合。     

合理利用空间信息的MRI脑部图像分割

在合理利用空间信息的基础上,提出了一种更准确,紧致性和分离性更好的分割算法。该算法首先定义一个空间函数,并在其中引入一个控制参数,该参数可以对噪声点、边缘点以及区域内部的点进行区别对待,然后用空间信息更新隶属度。实验结果表明,该算法效果要明显优于sFCMpq算法及其改进算法(EsFCMpq)。     

Defect detection using feature point matching for non-repetitive patterned images

Defect detection is an important technology for the quality control in the production process of wafer, TFT-LCD and PCB. Inspection is performed using the finished product’s image. The images are classified into two different groups—images with a repetitive pattern on a regular cycle and images without a repetitive pattern. A standard object for comparison is required, because manual defect detection is not possible for areas without repetitive patterns. In such areas, defect detection occurs through contrasting a reference pattern to the pattern being inspected. Methods of inspection using reference image have been researched but have limitations due to their requirement of precise alignment of the images. This paper proposes a method of defect detection to overcome such limitation using feature point matching. Feature points are extracted using a corner detector and detects defect by finding a correspondence between two feature point sets. Performance of the proposed method is evaluated by using Wafer SEM images and compared with conventional methods. Experiment results demonstrate the proposed method achieves higher detection accuracy than conventional methods and is less sensitive to alignment error and noise.     

基于灰度共生矩阵和聚类方法的木材缺陷识别

文章提出了一种新的基于灰度共生矩阵的木材纹理特征提取和聚类分析的木材缺陷识别方法。该方法基于灰度共生矩阵,提取5个具有代表性的纹理特征：能量（E）,惯性矩（I）,均值和（SOA）,聚类阴影（SOC）,方差和（SOV）,实现数据降维,对产生的特征数据集分别利用k-means算法及AP算法进行聚类,自动找出并正确标识木材表面缺陷位置,并对比分析不同木材表面缺陷类型识别效率。实验结果表明,该方法能快速有效地进行木材表面缺陷自动识别。     

基于谱系聚类的粗糙集离散化方法在胶合板缺陷检测中的应用

采用谱系聚类方法对大量数据进行离散化处理,通过散点图和树状图自动寻求最优分类数目。胶合板缺陷检测样本数据离散化实验结果表明,该方法根据数据内部结构特点,进行了可变的类别划分,与可k均值聚类和模糊聚类相比,该方法使约简的条件属性进一步减少,有利于数据的后期处理,并且进一步减少了系统的运行时间。     

扫描激光线源技术检测金属表面缺陷位置实验研究

利用扫描激光线源(Scaing Laser Line Source,SLLS)技术,提出一种金属表面缺陷位置检测方法,并搭建实验平台,实验研究该检测方法对金属表面缺陷位置的检测效果。方法中利用扫描激光线源技术对样品表面进行扫查,在样品内部激发超声信号,采用双波混合干涉方法实现对激发信号的探测,根据激发的瑞利波和反射回波与缺陷位置的关系,确定激发点和探测点到缺陷的距离,从而确定缺陷的位置。实验研究中分别以线源位置和探测点位置为基点确定缺陷位置,结果表明以探测点位置确定缺陷位置时,定位相对误差为0.23%。     

How repeated data points affect bug prediction performance: A case study

In defect prediction studies, open-source and real-world defect data sets are frequently used. The quality of these data sets is one of the main factors affecting the validity of defect prediction methods. One of the issues is repeated data points in defect prediction data sets. The main goal of the paper is to explore how low-level metrics are derived. This paper also presents a cleansing algorithm that removes repeated data points from defect data sets. The method was applied on 20 data sets, including five open source sets, and area under the curve (AUC) and precision performance parameters have been improved by 4.05% and 6.7%, respectively. In addition, this work discusses how static code metrics should be used in bug prediction. The study provides tips to obtain better defect prediction results.     

Recognition of dynamic visual images based on group transformations

A new dynamic image recognition method for image sequences based on two-level recognition of individual states and a set of states as a whole is considered in the paper. Integrated and invariant estimations of internal automorphisms are suggested. The modified collective decision-making method for dynamic image recognition uses four types of pseudodistances, in order to obtain a measure of similarity of input dynamic images with dynamic reference images depending on representation of dynamic features, as sets of numerical parameters, sets of vectors, or sets of functions. The experimental data on dynamic image recognition based on various measures of similarity of the sample with the reference pattern, which take into account isomorphic and permissible homomorphic transformations of visual projections, are presented.     

Application of random sets to model uncertainty of road polygons extracted from airborne laser points

High point densities obtained by today’s laser scanning systems enable the extraction of features that are traditionally mapped by photogrammetry or land surveying. While significant progress has been made in the extraction of roads from dense point clouds, little research has been performed on modelling uncertainty in extracted road polygons. In this paper random sets are used to model this uncertainty. Based on the accuracy reported by the data provider, positional errors in laser points are simulated first by a Markov Chain Monte Carlo method. An algorithm is developed next to detect the positions of road polygons in the simulated data and integrating the random sets for the uncertainty modelling. This algorithm is adapted to point data with different densities and variable distributions. Uncertainty modelling includes modelling of the dependence between the vertices of a road polygon. Road polygons constructed from vertices with different truncated normal distributions along with their uncertain line segments are represented by random sets, and their parameters are estimated. The effect of distributions on the area of the mean set is analysed and validated by a set of reference data collected from GPS measurements and image digitising. Results show that random sets provide useful spatial information on uncertainties using their basic parameters like the core, mean and support set. The study shows that random sets are well-suited to model the uncertainty of road polygons extracted from point data.     

基于小波与数学形态学的木材缺陷检测

木材缺陷检测是木材加工中的重要步骤,为了实现木材缺陷自动检测,提出了一种基于小波与数学形态学的缺陷检测方法。首先用多尺度小波对缺陷图像进行分解,滤除缺陷图像中的干扰信息,然后进行小波重构,在重构图像上进行形态学bottom-hat变换,结合阈值处理和区域生长检测出各种木材缺陷。实验表明,该方法具有高效准确的特点,能够满足木材加工过程缺陷检测的实际需求。     

Singular point detection by shape analysis of directional fields in fingerprints

This paper presents a new fingerprint singular point detection method that is type-distinguishable and applicable to various fingerprint images regardless of their resolutions. The proposed method detects singular points by analyzing the shapes of the local directional fields of a fingerprint image. Using the predefined rules, all types of singular points (upper core, lower core, and delta points) can be extracted accurately and delineated in terms of the type of singular points. In case of arch-type fingerprints there exists no singular point, but reference points for arch-type fingerprints are required to be detected for registration. Therefore, we propose a new reference point detection method for arch-type fingerprints as well. The result of the experiments on the two public databases (FVC2000 2a, FVC2002 2a) with different resolutions demonstrates that the proposed method has high accuracy in locating each types of singular points and detecting the reference points of arch-type fingerprints without regard to their image resolutions.     

改进C-V模型的木材缺陷彩色图像分割研究

分析了木材节子缺陷图像的特点,将彩色图像作为一个整体的图像进行处理,保护了彩色图像信息的特性,提出了一种基于AOS的扩展C-V矢量模型及背景填充耦合的木材节子缺陷彩色图像分割算法。对Chan-Vese提出的基于Mumford-Shah模型的水平集矢量图像分割模型进行了改进,使分割速度得到了提高;用AOS算法改进了原模型的差分格式,使得差分格式无条件稳定;结合背景填充技术,使得到的新图像缩减了目标与背景间的特征差别。实验结果表明该方法可以较好地实现对木材死节、活节和虫眼等缺陷的彩色图像分割,也可实现对单板多节子缺陷彩色图像的分割,为木材缺陷边缘检测提供一种行之有效的方法。