An Iterative Deep Homography Network Based on Correlation Content Calculation for High-Precision Image Registration of Fly-Capture Imaging System

作     者：Fang, Qiu Gao, Changqing Jiang, Tianjian Zhou, Xianen Wang, Yaonan 

作者机构：Hunan University National Engineering Research Center for Robot Visual Perception and Control Technology School of Electrical and Information Engineering Hunan Changsha 410082 China Department of Jiangxi Communication Terminal Industry Technology Research Institute Co. Ltd China 

出 版 物：《IEEE Transactions on Instrumentation and Measurement》 (IEEE Trans. Instrum. Meas.)

年 卷 期：2025年第74卷

学科分类：0808[工学-电气工程] 0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 0804[工学-仪器科学与技术] 0703[理学-化学] 

基　　金：Major Project of Science and Technology of China Strategic Research and Consulting Project of Hunan Institute of China Engineering Technology Development Strategy Special Funding Support for the Construction of Innovative Provinces in Hunan Province National Natural Science Foundation of China Hunan Provincial Department of Education Scientific Research Project Major Project of Yuelushan Industrial Innovation Center Hunan Provincial Central Leading Local Science and Technology Development Fund Project 

主　　题：deep learning Fly-capture imaging system Homograph matrix Microwave components Registration algorithms 

摘      要：This article presents a novel iterative deep homography network based on correlation content calculation (IDHNCCC) for high-precision image registration of a fly-capture imaging system. The new high-precision fly-capture imaging system (HFI) is designed to obtain high-resolution images of radar microwave components (RMC) and improve surface defect detection efficiency and autonomy while reducing hardware costs. The HFI consists of four subsystems: image acquisition mechanism, control system, image registration algorithm, and software system, which could achieve clear capture and precise imaging of RMC at high speed. The system adopts the IDHN-CCC algorithm to reduce the impact of motion errors on imaging stitching quality, which does not rely on precise camera motion positioning and coordinate transformation to achieve high-precision registration of multiple continuous shooting images. Experimental studies demonstrated the advantages of the IDHN-CCC-based HFI, which can achieve RMC images in low overlap rate fly-capture scenes with high registration accuracy. © 1963-2012 IEEE.