One-dimensional-based automatic defect inspection of multiple patterned TFT-LCD panels using Fourier image reconstruction

Thin film transistor-liquid crystal display (TFT-LCD) has been used in a wide range of electronic devices. For large-sized and high-density TFT-LCD panel inspection, a high-resolution line scan is demanded. A TFT-LCD panel image at a fine resolution presents very complicated patterns with less regularity. The paper proposes a non-referential defect detection scheme that directly works on the one-dimensional (1D) line images using the Fourier image reconstruction. The 1D grey-level line image is first divided into small segments, each of the length of the repeated period for a given TFT-LCD panel. The divided segments are then combined as a two-dimensional (2D) image. The frequency components corresponding to the 1D background pattern can be easily identified in the 2D Fourier spectrum. By eliminating the frequency components in the 2D Fourier spectrum that represent the periodic structural pattern of the combined 2D image and then back-transforming the image using the inverse Fourier transform, the 2D Fourier reconstruction process can effectively remove the complicated background pattern and well preserve local anomalies. Experimental results on a number of micro-defects embedded in different patterned regions of TFT-LCD panels show that the proposed method can reliably detect various ill-defined defects.     

面向视觉检测的深度学习图像分类网络及在零部件质量检测中应用

基于深度学习图像分类是视觉检测应用的基本任务。该文系统总结基于模型深度化图像分类网络、基于模型轻量化图像分类网络及其他优化网络主要思想、网络结构、实现技术、技术指标、应用场景,指出网络模型深度化、轻量化分别有助于提高图像分类准确性、实时性。最后,面向零部件质量检测需求,应根据其类型多少、结构复杂程度、特征异同等特点,结合实时性要求,选择合适的图像分类网络构建零部件质量智能检测系统。     

Selection of inspector for visual inspection

This paper presents a statistical method for selecting the best inspector for visual inspection having the highest defect-finding ability among k(≥2) prospective inspectors.

A sample containing known number of defects are submitted to each of k prospective inspectors. As the sample size is increased, sampling variation decreases so that we have a better chance of selecting the truly best inspector, but at the same time the cost of sampling is also increased, The problem is how to determine the sample size required for selecting the best inspector with specified probability P*.

This problem is formulated from the indifference zone point of view with an application of the loss function which takes into consideration the opportunity cost of making a wrong selection and the cost of sampling. Minimax solution is obtained under the restriction that the best inspector is to be selected with the specified probability P*. The selection procedure is applied to the visual inspection of car body painting.     

Improving visual inspection using binocular rivalry

In this study, we solve the non-identical parallel CNC machine scheduling problem. We have two objectives: minimizing the manufacturing cost (comprising machining, non-machining and tooling costs) and minimizing the total weighted tardiness. The tooling constraints affect the non-machining times as well as the machining conditions, such as cutting speed and feed rate, which in turn specify the machining times and tool lives. We propose a two-stage algorithm to find optimal machining conditions and to determine machine allocation, tool allocation and part scheduling decisions. The proposed algorithm generates different schedules according to the relative importance of the objectives.     

基于图像分析技术的螺纹自动化测量系统研究

为提高紧固件螺纹检验效率,实现经济的螺纹全参数检验,通过工业相机和光源选型,标定方法设计搭建了螺纹图像摄取系统,通过算法设计和优化编制了配套软件,对图像分析技术应用于螺纹检测分析进行了研究,并对测量系统进行了不确定度评定。研究表明基于图像分析技术的螺纹自动化测量系统可实现。     

自适应图像增强的管道机器人缺陷检测方法

针对管道检测过程中图像采集光照不均匀、缺陷边缘提取不准确的问题, 提出一种基于自适应图像增强的管道机器人缺陷检测方法。首先设计单尺度Retinex自适应图像增强算法, 利用引导滤波对图像进行照度分量估计, 经自适应Gamma矫正得到光照均衡图像, 实现自适应图像增强;再对传统Canny边缘检测方法进行改进, 采用双边滤波平滑图像, 通过迭代阈值法进行缺陷图像分割, 根据边缘像素相似性进行连接, 实现缺陷轮廓的有效提取。搭建基于自适应图像增强的管道机器人缺陷检测系统, 利用履带式小车搭载云台摄像机, 对管道内壁缺陷进行全方位视觉检测。实验结果表明, 本文的检测方法可自适应矫正图像亮度, 图像亮度不均匀明显改善, 相比次优算法, 图像信息熵提升2.4%, 图像平均梯度提升2.3%, 峰值信噪比提升4.4%, 可有效提取出管道缺陷边缘, 缺陷识别准确率达到97%。     

Illumination compensation methods for automatic fiber connector inspection

Unevenly distributed image illumination is a problem in an automatic machine visual inspection system used for fiber connector defect detection. This unevenly distributed illumination makes it difficult to find an appropriate horizontal threshold to convert a gray-scale image to a binary image to separate surface defects, such as blobs and scratches, from the background of the connector. This paper proposes some effective methods to compensate the unevenly distributed illumination problem so that standard image processing methods can be used effectively to detect the defects of the fiber connectors.