基于机器视觉的芯片图像快速定位及校正

Fast Localization and Correction of Chip Images Using Machine Vision

芯片精确定位是集成电路制造与生产的一个重要环节。针对芯片定位速度慢、精度低等问题,研究一种基于OpenCV视觉库的芯片图像快速定位及校正方法。首先对工业相机获取的芯片图像进行预处理,其次采用归一化互相关(Normalized Cross Correlation,NCC)方法对待检测图像进行模板匹配,最后使用仿射变换校正芯片图像。为加快匹配速度,采用图像金字塔采样芯片图像,经测试得出最佳金字塔层数为4。实验结果表明,所提方法能够准确地定位和校正芯片图像,为提高生产效率和检测质量提供了一种有效的解决方案。

Accurate positioning of chips is an important part of integrated circuit manufacturing and production.To address the issues of slow chip positioning speed and low accuracy,a fast chip image positioning and correction method based on OpenCV visual library is studied.Firstly,preprocess the chip image obtained from the industrial camera,then use the Normalized Cross Correlation(NCC)method to match the template of the detected image,and finally use affine transformation to correct the chip image.In order to improve the matching speed,an image pyramid was used to sample the chip image,and after testing,the optimal number of pyramid layers was determined to be 4.The experimental results show that this method can accurately locate and correct chip images,providing an effective solution for improving production efficiency and quality inspection.