With the rapid development of integrated circuit manufacturing technology, in advanced packages such as wafer-level packaging, the package size is getting smaller and smaller, and the defect detection resolution requirements have been raised from micron level to sub-micron level. The study of submicron-level 2D and 3D defect detection and complex defect identification and classification methods can provide key technical support for the research and development of semiconductor chip packaging defect detection system with independent intellectual property rights, and then lay the foundation for solving the stuck neck problem faced by the semiconductor sealing and testing industry and improving the yield of devices. Basics.

Recently, Professor Xia Haojie's team proposed an image super-resolution reconstruction network based on convolutional neural network for small targets by combining image super-resolution reconstruction with target detection and measurement, and designed and built a corresponding experimental system. This method breaks through the resolution limitation of the original hardware platform, effectively improves the imaging quality of small targets, and significantly improves the measurement accuracy of targets on the basis of the original hardware platform.

The relevant research results are entitled Channel Attention Network and System Design for Small Target Measurement as the cover article in Optical and Precision Engineering (EI, Scopus, Chinese Core Journals, 《Classification Catalogue of High-quality Science and Technology Journals in the Field of Instrumentation》 and 《High-quality Science and Optical Engineering》 Published in the T1 journal）. At the same time, the official WeChat platform of China Optics pushed the achievement. The first author of the article is Fu Yangwei, a graduate student, and Professor Zhang Jin is the corresponding author of this article. The research is funded by projects such as the National Natural Science Foundation of China.