Flexible IC substrate surface defect hierarchical classification method based on neural network

Abstract

The invention discloses a flexible IC substrate surface defect hierarchical classification method based on a neural network, and the method comprises the following steps: constructing a flexible IC substrate defect hierarchical classification tree model based on the neural network, which comprises a root node, a branch node and a leaf node; the root node performs identity identification on different batches of IC substrates and selectively activates a corresponding branch node model; the branch nodes correspond to different batches, each branch node model performs binary classification on the health state of the IC substrate through a data set organization strategy of balanced probability distribution in combination with a support vector machine, and leaf nodes cooperate with root nodes to classify output defect data and trace defect positions; and utilizing a YOLOv3 detector based on multi-scale feature dense pyramid connection to execute defect classification and position traceability. According to the method, multi-batch identity recognition and defect type rapid classification and positioning of the IC substrate are realized, and the learning performance and defect detection accuracy of a deep learning model under an unbalanced data set are improved.

Description

technical field [0001] The invention relates to the technical field of defect identification of high-density flexible IC substrates, in particular to a layered classification method for surface defects of flexible IC substrates based on neural networks. Background technique [0002] High-density flexible IC substrates are widely used in electronic products with miniaturized, lightweight and movable characteristics. The integrity and reliability of electronic product performance largely depend on the quality of flexible substrates. Different defects on the surface of the IC substrate will lead to product dysfunction to varying degrees, and even serious products will be directly scrapped. [0003] With the rapid development of the IC industry, the increasing demand for dense and miniaturized packaging of high-end chips has made the manufacturing process of the matching high-density chip packaging substrates more and more complicated and cumbersome, resulting in the risk of def...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the defects and deficiencies in the prior art, the present invention provides a neural network-based layered classification method for surface defects of flexible IC substrates, which solves the problem of automatic identification of multiple batches of substrates and corresponding batches of defects in the quality control process of high-density IC substrates. Intelligent and fast classification and positioning of multiple defect types for the first time, while improving the learning performance of the deep learning model under unbalanced data sets, and achieving higher defect detection accuracy

Images