中文摘要：

柔性印刷线路板中LED的粘贴需经过点银、贴片、封胶等工艺，其中贴片过程中会产生漏粘、异物、粘斜、偏移等一系列质量问题；提出了一套基于机器视觉的贴片质量在线检测方法：通过最小外接矩形获取有效区域的范围，利用基于轮廓特征的模板匹配对贴片所在连接盘进行定位；利用动态模板匹配检测漏粘缺陷；通过平均灰度值比较法检测LED上的异物缺陷；采用有效区域与LED区域的交集与差集对粘斜与偏移缺陷进行检测；该方案已应用于工业生产线，并且具有较高的检测准确率。

英文摘要：

The LED on the flexible printed circuit board need to be gone through silver point, patch, glue sealing and other processes. During the patch, there could produce a series quality problem, such as patch leak, rarity, stick inclined and offset and so on. Aiming at a bove--mentioned flaws, an online detection method of patch is put forward by machine vision： Smallest bounding rectangle is adopted to get the range of effective area, and template matching based on contour--feature is used to locate the land of patch. Dynamic template matching is used to detect patch leak. Average gray value is adopted to detect rarity. The intersection and difference of effective area and LED domain is applied, which detect the stick inclined and offset. The proposed scheme which has been already used on--line indicated that this scheme has a high detection accuracy.