欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2020, Vol. 37 ›› Issue (2): 234-241.DOI: 10.7523/j.issn.2095-6134.2020.02.013

• 多相流专栏 • 上一篇    下一篇

基于高斯回归预测的超声成像高分辨率重建

刘皓, 谭超, 董峰   

  1. 天津大学电气自动化与信息工程学院 天津市过程检测与控制重点实验室 天津 300072
  • 收稿日期:2019-02-12 修回日期:2019-04-30 发布日期:2020-03-15
  • 通讯作者: 董峰
  • 基金资助:
    国家自然科学基金(51976137)和天津市自然科学基金(19JCZDJC38900)资助

High resolution reconstruction of ultrasonic tomography based on Gaussian regression prediction

LIU Hao, TAN Chao, DONG Feng   

  1. Tianjin Key Laboratory of Process Measurement and Control, School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
  • Received:2019-02-12 Revised:2019-04-30 Published:2020-03-15

摘要: 油水两相流过程的可视化检测对多相流研究与石油和化工等工业生产具有重要意义。超声层析成像技术具有非侵入、无辐射、安装方便等优点,在油水两相流检测中有很好的应用前景。针对超声层析成像实现低对比度声学特性的油、水两相介质分布重建问题,提出一种高分辨率图像重建方法。该方法通过构建基于阶段奇异值分解的降维同步代数重建算法,在保证重建精度的同时提高重建速度,并采用高斯回归模型将低分辨率的重建结果转化为高分辨率图像。仿真模拟测试结果表明,该算法重建图像误差小、分辨率较高,能满足实时成像要求。

关键词: 油水两相流, 超声成像, 截断奇异值分解, 高斯回归模型, 同步代数重建

Abstract: Visualized measurement of oil-water two-phase flow is of vital importance in multi-phase flow measurement and process industry. Ultrasonic tomography (UT) is widely applied in two-phase flow measurement due to its non-invasive, non-radiative, safe, and portable advantages. Aiming at UT reconstruction of oil-water two-phase medium with low acoustic impedance, we propose a high-resolution reconstruction algorithm based on Gaussian regression prediction. The iteratively simultaneously algebraic reconstruction technique is modified with high pass filter, and the truncated singular value decomposition is used to reduce dimensionality of the coefficient matrix. Gaussian process regression is applied to gain high-resolution image. Simulation results indicate that the proposed algorithm provides high accuracy and high-resolution reconstruction with fast computing speed.

Key words: oil-water two-phase flow, ultrasonic tomography, truncated singular value decomposition, Gaussian process regression, simultaneous algebraic reconstruction technique

中图分类号: