【作者】 刘艳；

【导师】 赵洪；

【作者基本信息】 哈尔滨理工大学 ， 测试计量技术及仪器， 2009， 博士

【摘要】 透明及半透明聚合物薄膜或基片材料,如X光照相用聚酯基片、电影胶片、液晶显示(liquid crystal display,LCD)面板基片、高级包装或印刷用聚合物膜等材料中,如果存在缺陷就会影响材料的使用性能和销售价格。目前的工业生产中,聚合物薄膜或基片材料是以压延、流延、吹膜等形式生产的,造成的缺陷常见形式有凝胶物、外来杂质、表面划伤等,出现的形态多为颗粒状或不规则形状。研究一种高速高分辨力的聚合物薄膜或基片材料缺陷检测系统对制造过程中的产品进行100%的在线检测,对于全面控制产品的质量是十分有意义的。本文对CCD扫描式聚合物薄膜缺陷检测的关键技术进行了深入研究。设计了聚合物薄膜缺陷实时检测系统,分析了镜头的低通滤波效应,用平均值采样原理分析了CCD信号的空间频域响应和时间频域响应,根据CCD采样造成信号频谱延拓的特点获得了理想情况下的滤波器截止频率,并获知了平均值采样对图像高频信息的影响。分析了低通滤波器不同截止频率对图像高频信息及测量分辨力的影响,表明了可调参数滤波器对于优化系统分辨力的重要意义。分析了CCD器件输出信号与光积分时间、CCD驱动频率和后续低通滤波器之间的关系。为了最大限度地保留聚合物薄膜缺陷图像的信息以及获得最优的信号信噪比,确定了后续信号处理方案,在现场可编程门阵列(field programmable gate array,FPGA)芯片内部设计了可调参数的CCD驱动和有限冲激响应(finite impulse response,FIR)滤波单元,满足了系统的高速实时性要求。计算机可以通过USB2.0总线对CCD的光积分时间、驱动频率和滤波器指标进行配置,满足了系统的灵活性要求,因此这种参数可调的CCD信号采集系统可以广泛的应用于不同的CCD扫描式图像检测系统中。根据聚合物薄膜缺陷检测中感兴趣区域即缺陷区域在整幅图像中出现概率极低的特点,以及缺陷区域图像与背景有较高灰度差的特点,用图像灰度均值代表背景灰度,用FPGA高速实时的提取出感兴趣区域的图像信息,在保留有用信息的前提下极大地压缩了图像数据。并且设计了背景灰度不均匀条件下动态提取缺陷图像的方法。在FPGA芯片内部设计了阈值计算单元、状态机单元等具体实现了图像压缩和数据封装,计算机根据USB2.0总线上传的数据重建了聚合物薄膜中的缺陷图像。为了进一步的压缩处理数据,研究了图像处理领域常用的图像矩阵算子,根据矩阵算子处理数据量大、算法简单、具有并行结构的特点,在FPGA芯片上实现了基于矩阵算子的图像预处理,为海量图像数据的在线处理提供了解决方案。通过实验证明,本文设计的聚合物薄膜缺陷检测系统能够实时准确地检测出聚合物薄膜或基片材料中缺陷的尺寸及灰度信息。更多还原

【Abstract】 The performance and selling price of transparent and semi-transparent polymer film or substrate materials, such as polyester substrate using for X-ray film, cine film, liquid crystal display(LCD) plate substrate, printing polymer and well-packed film, will be affected by the internal defects or the surface defects. At present the polymer film or substrate materials are always produced by rolling, casting, blowing and other methods. The common form of defects is as gel, extrinsic contaminants, surface scratch, etc. And the defects are always granular or irregular shape. Therefore, it makes sense to design a high speed and high resolution detection system for 100% online detecting and comprehensive controlling the product quality.Key technologies of CCD scanning polymer film materials defects detection are deeply investigated in this thesis. A polymer film materials defects detection system is established. The low-pass filter effects of lens are analyzed. Time and spatial frequency domain responses of CCD signal are analyzed using the average sampling principle. The filter cutoff frequency in an ideal condition is obtained according to the characteristics of signal frequency spectrum extension caused by CCD sampling. At the same time, the image high frequency changes caused by the average sampling are got. Also the detection resolution and image high frequency changes affected by different cutoff frequency of low-pass filter are analyzed, which shows the significance of adjustable parameters filter to optimizing system resolution. Relationship of the CCD output signal and optical integration time, and relationship of CCD driving frequency are described. In order to obtain the most defects image information, as well as the optimal signal-noise ratio, a follow-up signal processing scheme is established. The CCD driving and finite impulse response(FIR) filter units with adjustable parameters are designed on one field programmable gate array(FPGA) chip, satisfying the high speed and real-time requirements of the system. A computer may configure the CCD integration time, CCD driving frequency and the filter parameters through USB2.0 bus, which make the detecting system flexible. So this CCD signal acquisition system with adjustable parameters can be widely used in different CCD scanning image detecting systems.There are two outstanding characteristics in polymer defects detection system. One is that region of interest(defect region) in a whole image is a small probability event. The other is big gray scale difference between defect region and background image. According to the two characteristics, the image gray scale is set to represent background gray scale and the region of interest image is picked up by FPGA in real time. Thus the image data is greatly compressed on the premise of retaining useful information. A dynamic extraction image method is designed under uneven background gray scale conditions. A threshold calculation unit and state machines are set on the FPGA to implement image compression and data packages. And the computer reconstructs the defects images according to the uploaded data through USB2.0 bus. Commonly used image matrix operators in the image processing field are studied in order to further compress data. Image pre-processing technology is achieved on FPGA chip according to the characteristics of the matrix operators such as large amounts of data, simple arithmetic and parallel structures. This method provides a solution for the massive image processing data. The experiments proved the polymer film materials defects detection system presented here can accurately detect defects in size and gray information in real-time.更多还原