基于局部马尔科夫模式的FPC薄浆缺陷检测

FPC trace paste attenuation detection based on local Markov pattern

薄浆是柔性印刷电路板（FPC）导电涂层的一种缺陷,为避免其对电路性能产生不良影响而造成经济损失,检测薄浆位置从而进行修补或直接剔除板材均是合理的处理方式。FPC线路结构复杂,同时由于其线路涂层结构特性和采集环境因素影响,所获得的图像容易出现＂斑点＂噪声、光照不均匀等现象,这些因素导致了当前算法检测性能降低。针对该问题,提出了一种基于局部马尔科夫模式（LMP）的检测方案。首先给出LMP算法的基本模型,然后将预处理后FPC图像划分成大小均匀的区块,以降低光照影响;接着,提取各区块内部线路区域的直方图特征并以此为依据选取种子点,同时为各种子点赋予其作为缺陷的初始概率,进而计算其LMP值,并通过LMP的数值大小识别其中的薄浆缺陷像素。在自建图库SUT-F2上进行了测试,结果显示方法对薄浆缺陷检测的等误率仅为4.06%,相对于其他典型纹理特征提取和薄浆检测方法其等误率至少降低了5.14%,表明了方法的高效性,具有实际应用价值。

The paste attenuation mentioned in this paper is a conductive coating flaw on flexible printed circuit（FPC） board. For avoiding its harmful effect on potential economic losses,it is a reasonable way to locate the position of paste attenuation for repairing or getting rid of the unqualified board directly. Due to the complex circuit structure and the influence by the circuit coating structure characteristics and the factors of acquisition environment,‘speckle＇noise and nonuniform illumination phenomenon usually exist in the acquired images,which result in the decrease of current detection algorithm performance. Aiming at this issue,a detection scheme is proposed which is based on the local Markov pattern（LMP）. Firstly,the basic model of LMP algorithm is built. Then,the FPC image after preprocessing is divided into blocks with uniform size to reduce the impact of illumination. Finally,the histogram feature of circuit regions in the blocks is extracted,and the seed point is selected according to this. In the meantime,the seed points are assigned as initial probability of the flaws,and then its LMP value is calculated and the pixels of paste attenuation flaw are recognized according to the LMP numerical value. Results show that the paste attenuation flaw detection EER of suggested method is only 4. 06 percent on the testing of self-built image storage SUT-F2,which decreases at least 5. 14 percent comparing to the other typical methods of textural features extraction and paste attenuation flaw detection. The above facts prove the high efficiency and application value of the suggested method.