基于FGT-FBP重构算法的焊接缺陷检测

针对焊接过程产生的缺陷,提出一种磁光成像传感的模糊灰度变换和滤波反投影(FGT-FBP)重构检测方法. 研究焊接缺陷的几何特征,通过分析裂纹和未熔合两种不同焊接缺陷在交变磁场励磁下的磁光成像特征,设计模糊规则,对磁光图像进行模糊灰度变换. 增强磁光图像对比度,使焊接缺陷形态趋势可视化,实现描述磁光成像焊接缺陷细节的无参考型图像评估方法. 对FGT处理的焊接缺陷磁光图进行旋转投影,并经过快速傅里叶变换和改进的滤波器进行滤波去噪,消除伪影后进行反投影变换实现焊接缺陷图像的重构. 利用滤波反投影重构算法进行去噪,可有效突出焊接缺陷特征. 最后结合阈值分割和边缘检测实现焊接缺陷检测. 结果表明,该方法能较准确检测裂纹和未熔合两种焊接缺陷.     

固体火箭发动机高能X射线探伤胶片影像三维重建

固体火箭发动机是导弹武器系统的主要动力装置,需要定期进行无损检测。常规的检测方法是对X射线照相得到的二维胶片进行评判,然而利用二维图像来反映三维空间结构,在直观显示上存在很大的缺陷性。介绍了利用二维胶片图像还原发动机的空间三维结构的基本重建原理和六大体系结构,即数字图像的获取、各项参数的标定、图像的预处理和特征提取、图像的立体匹配以及图像信息三维重建。     

一种基于图像质心的焊缝跟踪新方法

研究一种基于图像质心识别的电弧焊焊缝跟踪新方法。通过视觉传感器获取焊接区熔池图像，抽取图像质心坐标并构成状态向量，建立一种基于图像质心的状态方程和位置测量方程。在此基础上，应用卡尔曼滤波对图像质心位置进行状态估计，在时域中采取递推计算的方式得到最小均方差条件下的焊缝位置最佳预测值．从而消除过程噪声和测量噪声引起的焊缝位置测量偏差。计算机仿真和实际焊接试验结果显示该方法可有效地提高焊缝跟踪精度。     

3D image reconstruction using cone beam tomography

3D X-ray cone beam tomography for possible NDT application is illustrated via the Feldkamp convolution backprojection reconstruction algorithm. Gauss and Ham filter functions are implemented in order to alleviate artifacts (streaks and ringing). In addition, the effect of cone vertex position error on image reconstruction is investigated. Finally, film based X-ray projection radiographs were scanned into the computer to reconstruct the 3D images and verify the performance of the software program.     

Enhancement of thermographic images as tool for structural analysis in earthquake engineering

In this paper, we present an application of a reconstruction method to thermographic images employed to analyze the response of a masonry structure under seismic actions.At first the theory of linear multivariate sampling Kantorovich operators is presented. By means of the above operators, we are able to reconstruct images taken from thermographic survey of masonry walls, and enhance their quality. Digital image processing of reconstructed images allows us to identify the mutual arrangement of the blocks (made of stones and/or bricks) and mortar joints inside the wall portion analyzed, and therefore the texture of the masonry. Subsequently, the texture has been used to estimate the equivalent elastic properties of the masonry by means of homogenization techniques. Finally a real-world case-study is analyzed, taking into account the mechanical properties estimated from reconstructed thermographic images and evaluating the structural response in terms of modal analysis.     

基于透红外视觉传感的GMA-AM熔池图像质量评价

针对熔化极气体保护电弧(gas metal arc,GMA)增材制造(additive manufacturing,AM)图像感知,提出一种透红外熔池视觉传感方法. 为客观评价熔池图像质量,综合图像灰度、纹理、形状和频谱等四类特征定义了熔池图像质量评价参数φ. 结果表明,φ值越大,图像质量越好. 透红外熔池图像质量评价参数φ远大于近红外窄带熔池图像,熔池更清晰,对比度更高. 相比800,850和930 nm等透红外滤光片,990 nm透红外滤光片能过滤大部分电弧连续谱和特征谱,获得的熔池信息量最大,对比度更明显,边缘更清晰,细节信息更丰富,熔池图像质量最佳,是最佳取像窗口.     

基于感兴趣区域和小波变换的图像融合算法

为了提高图像融合效率,保持良好的边缘支持度,以便更好完成图像重构,提出了基于小波变换的图像融合算法。利用小波变换对信号多方向特征提取的优势,将多维图像信号分析化繁为简,在融合规则制定中,权衡图像融合的清晰度和融合算法复杂度,对图像信号低频信号和高频信号分开处理,低频进行加权平均,而高频进行局部方差处理,最后进行小波逆变换得到重构图像。实验证明,该方法重构图像效果清晰,且信噪比高。     

A projection method for reconstructing X-ray images of arbitrary cross-section

This paper describes a new method for reconstructing cross-sectional images of an object at arbitrary height and arbitrary angled planes. This method, called the geometric projection method, can produce cross-sectional images for arbitrary planes using only those images acquired from the focal plane of an object. This paper develops the mathematical formulae for the image reconstruction of arbitrary height and arbitrary angled planes. Then, by means of the derived formulae, the cross-sectional images of arbitrary planes are realized via computer simulations. Finally, by using experimental results conducted on the solder joints of a ball grid array package, it is shown that cross-sectional images for arbitrary planes of the object can be reconstructed correctly by the geometric projection method.     

快速成形逆向工程扫描实体重构法研究

针对现有快速成形逆向重构研究所忽视的图像数据信息,提出一种扫描实体图像数据重构的方法。该方法基于三基准点对齐和八邻域法实现轮廓提取,利用图像误差点消除和图像精简有效改善重构数据质量,并快速利用B样条曲线插值拟合出较精确的轮廓曲线。以实例展示了模型的实用性及算法的合理性与有效性。     

基于独立分量分析的X射线图像增强方法

为了改善由于散射造成的X射线检测图像对比度低,被测工件细节不易显示等缺点。针对退化图像中散射的分离,提出一种盲参数估计分离散射算法,该算法在假定图像由独立散射光谱组成的基础上,通过对散射偏振分析建立盲分离模型,来实现X射线图像的增强。采用小波变换得到图像子带图像,对盲参数间接估计。理论分析和实验结果表明,该算法在抑制散射的基础上有效的提高了图像的对比度和清晰度。     

Computed tomography with image intensifier: potential use for nondestructive testing and imaging of small objects

An image intensifier based computed tomography scanner and a tube source of X-rays are used for nondestructive evaluation, imaging of small objects for archaeological studies and conservation of works of art and micro analysis of soft materials. It consists of a charge coupled device (CCD) camera and an acquisition board. The CCD camera and the acquisition board allow image processing, filtration and restoration. The object is irradiated by an X-ray tube with a conical collimator and rotated on 180°. Projections are obtained at various angles as cross sectional image slices. A reconstruction program written in pascal is able to give the reconstruction matrix of the linear attenuation coefficients, simulates the matrix and related tomography. The flux emitted by the tube is filtered using the appropriate filters at the chosen optimum energy and reasonable monochromacy is achieved for all the images. Although X-ray imaging is a potential tool for strongly attenuating materials, the images of weakly attenuating materials provide new information to know about their structure and also the foreign body for the image reconstruction at an optimum value. The image of the plastic material which contains the internal defect is studied thoroughly at the optimum value in order to image the small objects for nondestructive testing, archaeological studies and conservation of works of art. The images are analysed using the profile data showing the internal defect of the object to obtain information at the optimum value. At the optimum value and with the aid of the tomographic image, the size and location of the defect could be ascertained.     

远距离采集图像损伤区域三维重建仿真研究

远距离采集图像的损伤区域形状基属性形变和噪音难以消除,三维重建难度大,传统重建方法不能有效修复损伤区域.提出基于体像素插值的图像损伤区域三维重建方法,在远距离图像采集中形变不规则的前提下,利用产生奇异点的规律进行奇异点插值补偿,提取到图像仿真投射映像属性点的非齐次矩阵后,利用体像素补偿损伤区域奇异点,并通过边缘冲突原理进一步消除奇异点,获取到更加真实的远距离采集三维重建图像.实验结果证明:所提方法增强了远距离采集图像损伤区域三维重建的信息量.     

The influence of component inclination on surface finish evaluation using digital image processing

Many researchers have so far used Machine Vision and digital image processing for grabbing images of machined surfaces, improving their quality by pre-processing and then analysed them for evaluation of surface finish with a reasonable success. In the conventional mechanical stylus method used for roughness evaluation, many of the fundamental requirements are taken care of during measurement which includes alignment of component with the stylus pick up movement, tracing length, filter cut off length, etc. Practical use of Machine Vision for surface roughness estimation faces many challenges, as in this case only image is used for evaluation and not the component. For example, if the component is kept at an angle during imaging, there is a possibility of getting distorted information and the consistency of evaluation/quantification would become a problem. Therefore, there is a need to ensure that the measured surface is kept horizontal when the image is being taken. In this work, estimation of the surface roughness has been done and analysed using digital images of machined surfaces obtained by a Machine Vision system deliberately maintained at varying angles. The optical surface finish values (G_a) estimated in all such cases using Machine Vision approach are compared with that obtained using conventional stylus method (R_a). An artificial neural network (ANN) is trained and tested to arrive at the R_a values using the input obtained from the digital images of inclined surfaces which include optical roughness parameters estimated and angle of inclination of test parts. The experimental result indicates that the surface roughness could be estimated/predicted with a reasonable accuracy using Machine Vision and ANN, respectively. In addition, a shadow removal algorithm is used to improve the quality of the images of inclined surfaces and then the optical roughness parameter is estimated. All the results are compared with that obtained using stylus method and analysed in this paper.     

无人机影像的快速拼接及其误差

利用经过差分的DGPS数据和IMU(Inertial measurement unit)获取的外方位元素,经过影像匹配产生的同名点,获得同名点地理坐标,使各个模型在此基础上连接起来。经过影像重采样用地理坐标代替像素坐标,使拼接后的影像具有拍摄时的地理坐标。实验结果表明:平坦地区投影差较小,可以不进行投影误差改正,直接利用左右影像同名像点进行影像拼接;丘陵山区因为投影误差的影响,影像拼接的误差累积较快,此时应采用分带拼接,避免误差积累。拼接后的影像坐标与拼接精度达到了1-2.5万比例尺影像图的精度。     

Cross-sectional imaging of building elements by new non-linear tomosynthesis techniques using imaging plates and Co radiation

The quantitave measurement of the positions and diameters of steel reinforcements in buildings is based on the concept of non-linear backprojection (NLBP). Advantages are 3D-image reconstructions from a small number of views, fast numeric algorithms and the consideration of the statistical, spectral and other characteristics of the object. For the radiographic inspection a ⁶⁰Co source was used and both X-ray films and imaging plates for detection. The imaging plates were combined with special filter screens for contrast optimization. The digitized images from the plates have a high dynamic range and can be read into a computer without a photographic developing process. A numerical preprocessing procedure was developed as space filtering in different directions followed by a combination of the image angle components. A generalized Hough or Radon transformation and a statistical evaluation of the resulting image was employed. The special advantage of this method is the reconstruction of direction-selective structures.     

一种剥皮算法在工业CT图像分割中的应用

图像分割是工业CT三维重构的重要步骤之一,目前还没有一种有效分割各类图像的通用算法。针对同心圆层状结构工业CT图像的特点,提出一种剥皮图像分割算法。该算法首先检测图像的边缘,利用边界跟踪方法求出层边界,再对层边界图像进行剥皮处理,然后用所得到的模型去对原始图像掩模,最后得到的图像即为最终分割图像。该方法对层状结构的工业CT图像具有较好的分割效果。     

基于视觉传感的焊缝中心检测技术研究

通过视觉传感器获取焊缝图像，采用多种去噪技术对焊缝图像中的噪声进行滤波试验。针对焯缝图像中的干扰噪声特性，对各种滤波方法进行了分析和比较，并对滤波效果较好的焊缝图像进行边缘检测计算。提出3种识别焊缝特征的组合图像处理方法，并研究提取焊缝中心特征信息的效果和精度。试验结果表明，对焊缝图像采用Wiener滤波与Robert边缘检测算法能较好地提取焊缝中心位置的特征信息。     

Segmentation of thermal images for non-destructive evaluation of bridge decks

The objective of the project presented in this paper is to automate the detection of subsurface defects in concrete bridge decks using infrared thermography. The algorithm developed for this purpose is based on the region growing approach, which segments the image and identifies the voids without human interference or prior knowledge of the conditions. The segmentation algorithm starts with the hottest pixels in the image as seed points, and then regions are grown around them based on a neighborhood selection criterion. The algorithm was tested on images collected from concrete bridge deck specimens containing various man-made defects and also on a defect-free control model. The experimental work successfully identified defects in concrete bridge decks up to 3 in. below surface using thermograph imaging.     

Enhancement algorithm for underwater weld seam image

It is hard to treat the underwater weld seam images for the reason of bad brightness, low contrast and less welding seam information, so a new enhancement algorithm is proposed here. Firstly, the high frequency component was separated by Gaussian filter from origin image, and then it is processed by improved local contrast enhancement(LCE) algorithm to enhance the edge information. Secondly, the gamma transform with adaptive parameters was used to strengthen the image brightness, furthermore, contrast limited adaptive histogram equalization(CLAHE) algorithm was applied to enhance the image contrast. Finally, the two manipulated images were integrated together to obtain the desired image. Experiments on typical images were carried out, and evaluation results showed that this designed algorithm can effectively improve image contrast, highlight welding seam information. Moreover, the image average grey value was moderate, and the information entropy and average gradient were much higher than other algorithms.     

基于小波分解和异步重构的图像信息增强算法

针对传统基于小波变化的图像信息增强问题,提出了一种结合小波分解和异步重构的图像信息增强算法.其中的异步重构方法并不采用通常的顺序,而是采用排序编码来执行.通过模拟人脑处理视觉信息的方式,该方法试图通过使用较少的频率通道来实现图像重构.首先,信号首先通过高斯差分滤波实现空间频率分解.然后选择最重要信息频率分量,每个像素利用最好的尺度进行重构,并且忽略弱响应尺度.实验结果表明:在局部重构的情况下,相比传统的标准排序重构算法,提出的算法能够快速的实现图像重建并达到较高的增强效果.