柔性印刷电路板缺陷检测系统研究

发布时间：2018-03-29 13:21

本文选题：柔性电路板　切入点：并行计算　出处：《华南理工大学》2015年硕士论文

【摘要】：近年来,柔性印刷电路板(Flexible Printed Circuit,FPC)被广泛应用于消费电子领域,随着科技的不断发展,制造工艺的不断提升,市场对柔性印刷电路板的产量和质量都提出了更高的要求。在实际生产中,自动光学检测设备被寄望能更快速、更准确、更稳定地找出流水线上的不良品,以达到提高生产效率和产线成品率的目标。适应现今计算机处理器往多核并行方向发展的趋势,本文创新地提出了一种基于并行比较的检测系统方案,并给出相应的机械结构及工作原理,力求大幅度提升柔性印刷电路板缺陷检测系统实时性能,同时控制设备的生产成本。文章把缺陷检测过程分为图像采集、图像预处理和检测算法三部分,同时在多核处理器和众核图形处理器两个平台下分别对图像预处理和检测算法进行并行优化加速。首先,通过引入USB转PCI-E板卡解决多路视觉在并行输入时的资源争用以及传输带宽不足的问题,利用多线程实现独立并行的图像采集功能。然后,分析图像预处理和检测算法的原理和工作特点,在多核处理器平台下设计了流水线处理和数据分解两种方案对程序进行加速,并对两种方式的优缺点进行了比较。最后,搭建CPU+GPU的异构计算平台,分析GPU的工作模式和结构特点,并基于CUDA(Compute Unified Device Architecture)的编程模型对图像预处理和检测算法进行加速。实验数据表明,并行处理的优化方式对提高检测系统的实时性效果显著,随着数据量的增加,在CPU+GPU的异构平台下加速效果更佳,可以达到1到2个数量级的速度提升。文中提出的并行对比检测方案与CPU+GPU的异构计算方式相结合,为改善柔性印刷电路板缺陷检测系统的实时性提供了一个新的实现思路。若对该检测方案深入研究并加以完善,有望用于实际的生产环境,极大地提升现有检测设备的工作效率。
[Abstract]:In recent years, flexible Printed circuit boards (FPCs) have been widely used in the field of consumer electronics. The market has put forward higher requirements for the output and quality of flexible printed circuit boards. In actual production, automatic optical inspection equipment is expected to be able to find out the bad products on pipeline more quickly, more accurately and more steadily. In order to achieve the goal of improving production efficiency and production line yield, and to adapt to the trend of multi-core parallel development of computer processors, this paper proposes a detection system scheme based on parallel comparison. The corresponding mechanical structure and working principle are given in order to greatly improve the real-time performance of the defect detection system of flexible printed circuit board and control the production cost of the equipment at the same time. Image preprocessing and detection algorithm are three parts. At the same time, the image preprocessing and detection algorithms are optimized and accelerated on the platform of multi-core processor and multi-core graphics processor respectively. By introducing USB to PCI-E board to solve the problem of resource contention and insufficient transmission bandwidth in parallel input of multi-channel vision, the independent and parallel image acquisition function is realized by multithreading. This paper analyzes the principle and working characteristics of image preprocessing and detection algorithm, designs two schemes of pipeline processing and data decomposition to accelerate the program under the multi-core processor platform, and compares the advantages and disadvantages of the two methods. The heterogeneous computing platform of CPU GPU is built, the working mode and structure characteristics of GPU are analyzed, and the image preprocessing and detection algorithm is accelerated based on CUDA(Compute Unified Device Architecture programming model. The experimental data show that, The optimization method of parallel processing has a remarkable effect on improving the real-time performance of the detection system. With the increase of data volume, the acceleration effect is better under the heterogeneous platform of CPU GPU. The parallel contrast detection scheme proposed in this paper is combined with the heterogeneous computing method of CPU GPU. This paper provides a new idea for improving the real-time performance of the defect detection system of flexible printed circuit board. If the detection scheme is deeply studied and perfected, it is expected to be used in the actual production environment. Greatly improve the efficiency of existing testing equipment.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN41;TP391.41

【参考文献】