Semiconductor chip gold thread defect classification method and system based on deep learning

Abstract

The invention provides a semiconductor chip gold thread defect classification method and system based on deep learning, and the method comprises the steps of shooting chips through a light field camera, obtaining central view images and depth information, and enabling each central view image to comprise two complete chips; segmenting the central view angle image to obtain a grey-scale image of the single chip; respectively marking outlines for gold threads of the grey-scale map of the single chip; performing defect classification on the grayscale image marked with the contour in combination with the depth information to obtain a data set; and performing gold thread defect classification on the semiconductor chip graph by using the data set. The invention is high in accuracy rate in test concentration, and three defects and intact category features of the gold thread can be effectively judged.

Description

technical field [0001] The invention relates to the fields of image processing and semiconductors, in particular to a deep learning-based method and system for classifying gold wire defects in semiconductor chips. Background technique [0002] With the rapid development of semiconductor technology and the wide application of integrated circuit IC chips, the packaging process of semiconductor chips has been severely challenged. In the face of the development trend of miniaturization of electronic products, the requirements for packaging technology are becoming more and more stringent. Semiconductor chips are connected to each other by the die and the pins of the chip, and some kind of high-conductivity metal wire is required to be used among them for connection. [0003] On the assembly line, after the die and pins and other parts are perfected, the wire bonding technology begins to play a role. Usually for different chips, the environment used will be different, and some ar...

AI Technical Summary

Problems solved by technology

These three types have high resolution and accuracy, but the cost is too high. Due to the need to take multiple images and process images of different depths, there are inevitable vibration errors in them, and there are also complicated scanning processes. The synthesis operation of the method leads to low detection efficiency, and has not been widely used in today's chip detection field.

Images