摘要现今科技有各种各样的封装技术。而在这不断发展的科学技术下，提出了一 种新型的封装技术——倒装芯片封装技术。这种技术在现在得到了广泛的使用， 但同时也存在着不足，芯片越来越小，各种新的要求的出现，封装出现错误的情 况越来越严重，错误的出现一定得有相应的检测，但现如今的检测方法和现在的 生产相比无法同步，所以对现在的检测方法进行发展和改进是非常有必要的。 72673
要对焊球缺陷进行智能检测首先要得到图像，而获取图像的同时不能对芯片 有所损伤，所以本文选择用超声波检测技术来检测焊点缺陷，这是无损检测技术 中的一种，这种技术实用性比较强，检测完后就得到了包含焊球信息的图片，再 提出了一个对扫描声图像智能识别的方法，用来对图像研究，包括以图像灰度值 梯度为阈值的分割算法，以数理统计为基础的特征提取等，然后运用 LVQ 神经网 络进行缺陷检测。
而为了实现焊球缺陷的智能检测，我们采用了基于 LVQ 神经网络进行焊球检 测的学习算法对那些超声波扫描后的图像进行分割，特征提取，然后运用算法进 行智能检测。
毕业论文关键词：超声波扫描 图像处理 焊球特征 LVQ 算法 焊球缺陷检测
Ball detection method based on neural network LVQ
Abstract Flip-chip packaging technology is a new packaging technology in the packaging field has been widely used。 However, due to the solder balls are hidden between the chip and the substrate, and therefore it is difficult to achieve an effective defect detection, and with the proposed downsizing, and the new requirements of the flip chip package failure situation is more serious。 And the existing flip chip detection methods are the existence of local deficiencies, too difficult to meet production needs, the development and improvement of flip chip defect detection method is of great significance。
Ultrasonic inspection techniques are solder joint defects in a non-destructive testing methods, and because of its practicality has been more widely used。 Through ultrasound scanning techniques were tested chip solder ball defect detection experiments, image information obtained after the solder balls, proposed a scanning acoustic image automatic identification method, conducted research on image processing techniques, including image gray value gradient threshold segmentation algorithms and mathematical statistics based feature extraction, and then use the LVQ neural network defect detection。
In order to achieve intelligent detection ball defects, we have adopted a learning algorithm based on neural network LVQ ball detecting the ultrasonic scan those images after segmentation, feature extraction, and then using intelligent detection algorithms。
Key Words: Ultrasound scan Image Processing Bump feature LVQ algorithm Bump defect detection
目 录
摘要-Ⅰ
Abstract--Ⅱ
图清单-Ⅳ
1 绪论 1
1。1 研究现状及意义 2
1。2 封装技术 3
1。3 缺陷检测 5
1。4 本章小结 5
2 SAM 焊球缺陷检测系统 6
2。1 SAM 原理 6
2。2 测试芯片 基于LVQ神经网络的焊球检测方法:http://www.youerw.com/jixie/lunwen_82707.html