Optimization and Validation of Wafer Surface Defect Detection Algorithm Based on RT-DETR

In response to the issue of poor detection performance on wafer surface defect spots and elongated scratches, an improved RT-DETR method for wafer surface defect detection is proposed. Firstly, a dynamic snake convolutional layer is introduced to detect elongated scratches where conventional convolu...

Full description

Saved in:
Bibliographic Details
Main Authors: Ao Xu, Yanwei Li, Hongbo Xie, Rui Yang, Jianjie Li, Jiaying Wang
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10892113/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In response to the issue of poor detection performance on wafer surface defect spots and elongated scratches, an improved RT-DETR method for wafer surface defect detection is proposed. Firstly, a dynamic snake convolutional layer is introduced to detect elongated scratches where conventional convolutional kernels fail to extract features effectively. Secondly, to address the problem of information loss in small targets, an attention-based Transformer encoder module and a feature fusion network based on residual thinking are proposed. Finally, verification is conducted using a wafer test dataset. Experimental results demonstrate that compared to the original RT-DETR method, the model exhibits a 4.1% improvement in detecting small particles and a 5.4% improvement in scratch detection performance. Fully meeting the requirements of intelligent manufacturing and high detection accuracy.
ISSN:2169-3536