Chip Surface Character Recognition Based on Improved LeNet-5 Convolutional Neural Network

摘要

It is very important to detect the correctness of character information on the surface of IC chip in the process of packaging and testing. In this paper, a character segmentation method based on rectangle contour is proposed and the LeNet-5 network model is optimized. The character segmentation algorithm based on contour can effectively segment the characters in both horizontal and vertical directions. The improved convolutional neural network model is verified on the surface character data set of IC chip and has achieved good results.