基于图像的PCB板的在线检测

【作者】 尹迎菊；

【导师】 罗大庸；

【作者基本信息】 中南大学 ， 计算机技术， 2008， 硕士

【摘要】 PCB在现代电子设备中占重要的地位,其质量直接影响产品的性能,目前PCB质量检测大都由人工目测完成,容易发生误检和漏检。因此PCB工业迫切需要在线自动检测。基于图像的PCB在线检测主要包括图像采集、图像的预处理、边缘检测、PCB的定位和瑕疵图像的检测识别几个环节。本文重点就以下几个问题作了具体研究:图像的预处理是在线检测的前提和基础,本文采用改进的中值滤波法和均值滤波法去除图像的噪声,基于拉普拉斯算法的图像的增强,采用最大类间方差法对PCB进行二值化处理;PCB图像的边缘检测研究中,本文采用基于全方位、多尺度的数学形态学法检测图像边缘;PCB的定位上,本文应用改进的Hough方法对PCB的圆形定位标志进行检测,并研究了具体的实现算法,通过实验证明此方法是比较好的算法;在PCB瑕疵识别中,基于多模板参考比较法的瑕疵检出方法,保留了参考比较法在速度上的优势,进一步提高了系统的检测能力,基于模板图像结构特征的句法结构瑕疵识别方法利用数学形态学理论提取模板图像的结构特征,并将这些结构特征与句法结构瑕疵识别方法相结合,有利于提高瑕疵图像识别速度和准确性,具有运算量小、计算速度快的特点。更多还原

【Abstract】 PCB is very important in modern electronic equipment, its quality has direct impact to the product function, At present, PCB quality inspecting is completed mostly from manpower visual measurement, and the mistake check up happened and misses. PCB industry needs the online automation detecting therefore urgently.Owing to that the image PCB checks system on line, mainly, the pretreatment, border detecting , the PCB allocation and blemish image detecting including image acquisition , the image distinguish several links. Priority has written the main body of a book with regard to several the following problems studying concretely:The image pretreatment is premise and basis checking system on line, The article uses middle value wave filtering law and mean value wave filtering improving to follow the noise dislodging an image, Algorithmic image of Laplace-based strengthening, Adopt maximal kind room method difference to follow the value carrying out two on PCB spending treatment; PCB image edge detecting studies middle , the main body of a book adopt the because of all-direction , many dimension mathematics morphological approach detecting image border to law; On the PCB allocation, Improvement Hough method the main body of a book is applied gives a test to circular PCB allocation sign, has studied and the concrete realization algorithm, good by the fact that the experiment testifies this method is to compare algorithm; In the PCB blemish distinguishing, the method based on multipy form board reference blemishes inspect up out method, have reserved the reference comparison method advantage on speed and highly improved the systematic detecting ability; the based on mul-board image structure characteristic syntax structure blemish distinguishes method uses mathematics morphology theory to abstract the form board image structure characteristic, and the combination between the structure characteristics and the syntax structure blemish distinguishing is beneficial to improving blemish pattern recognition speed and accuracy, having an arithmetic small, calculate the quick characteristic of speed.更多还原