Abstract:

In order to solve the problem of leakage and stickiness in the production of flexible printed circuit board LED, a set of on-line inspection method of LED stickiness quality based on machine vision is proposed. A large plot of 8,192×70,000 pixels is firstly pre-treated, and the rectangular sieve of zone is used to retrieve the rectangular sieve of zone unless the pads area by using the stencil and traversing the idea to get the pad profile of a single pixel as the positioning template. As asynchronous scale of template matching positioning LED bonding pad area, wick region roundness is calculated by region projection, according to roundness patch of different contrast adaptive parameter Canny edge detection for getting wick edge, the wick edge is fitted to obtain wick coordinates. Lastly, LED defect is judged by wick coordinates, and LED stick deviation is judged by calculating pad coordinates deviation and lamp core coordinates. Field test results show that accuracy of LED leakage detection is 100%, accuracy of stick slant LED detection is 97.3%, and average detection time is 1,583?ms, which proves that the proposed detection method has practical value.

Key words: LED paste quality, moment, asynchronous zoom, adaptive parameters, projection

摘要：

针对柔性印刷线路板LED在生产过程中出现的漏贴和粘偏缺陷问题，提出一套基于机器视觉的LED粘贴质量检测方法。对8?192×70?000像素的大图进行预处理，利用统计学中的矩求取区域的矩形度筛除非焊盘区域，通过模板并行遍历思想获得单一像素的焊盘轮廓作为定位模板。采用异步缩放模板匹配定位LED的焊盘区域，使用区域投影的方法计算灯芯区域的圆度，根据圆度对不同对比度的贴片进行自适应参数的Canny边缘检测得到灯芯的边缘，拟合灯芯边缘获得灯芯的坐标。通过灯芯的坐标判断LED的漏贴缺陷，通过计算焊盘坐标和灯芯坐标的偏差判断LED的粘偏缺陷。经过大量的现场测试，检测漏贴LED的准确率达到100%，检测粘偏LED的准确率达到97.3%，检测一张图片的平均时间为1?538?ms，证明该检测方法具有一定的实用价值。

关键词: LED粘贴质量, 矩, 异步缩放, 自适应参数, 投影法