基于小样本图像分类的芯片引脚缺陷检测方法

Research on automatic detection of pin defects of small sample SOP chips

为了检测SOP芯片的引脚缺陷,提出了基于小样本图像分类的SOP芯片引脚缺检测的方法。该方法包含了芯片定位识别、芯片倾斜矫正以及芯片引脚参数测量。根据芯片塑封体与视觉系统采集的样品背景的特征进行芯片定位识别算法研究。通过提取芯片塑封体的边缘点进行直线拟合并矫正倾斜的芯片算法研究。基于Lnent-5模型构建图像分类网络,以10×10的尺寸对预处理后的图像进行切割并分类,通过分类结果确定每个引脚的位置以及边界,并根据引脚边界的分类结果计算芯片引脚的关键尺寸,判断是否存在缺陷。实验结果表明总体检测率达到99%,此方法能够满足在小样本的情况下稳定、准确地检测出SOP芯片的引脚缺陷。

In order to detect the pin defect of SOP chip,a method of detecting the pin deficiency of SOP chip based on image classification was proposed.The method includes chip positioning identification,chip tilt correction and chip pin parameter measurement.According to the characteristics of the sample background collected by the chip packaging body and vision system,the chip localization and recognition algorithm was studied.By extracting the edge points of the chip molding body,the chip algorithm of linear fitting and alignment correction was studied.The image classification network was built based on Lnent-5 model,and the pre-processed images were cut and classified with the size of 10×10.The position and boundary of each pin were determined according to the classification results,and the key size of the chip pin was calculated according to the classification results of the pin boundary to determine whether there were defects.The experimental results showed that the overall detection rate reached 99%,and the sys-tem could meet the requirements of stable and accurate detection of pin defects of SOP chips in the case of small sam-ples.