Page 85 - 《应用声学》2023年第3期

第 42 卷 第 3 期 Vol. 42, No. 3
2023 年 5 月 Journal of Applied Acoustics May, 2023

⋄ 研究报告 ⋄

超声全矩阵数据联合稀疏重构的多测量

向量模型应用 ∗

严 凡 林 莉 金士杰 †

(大连理工大学无损检测研究所 大连 116085)
摘要：针对单测量向量模型等传统压缩感知方法处理超声全矩阵捕捉数据时，存在重构精度低和重构耗时长
等问题，该文研究了多测量向量模型应用的可行性。针对铝合金试块中不同深度的 Φ2mm 横通孔，分别使用
多测量向量模型中的多测量稀疏贝叶斯算法和单测量向量模型中的稀疏贝叶斯算法进行超声全矩阵数据重
构，并实施全聚焦成像。随后，引入归一化均方误差和阵列性能因子评价图像和信号的重构效果。实验结果表
明，稀疏贝叶斯算法在 25% 采样率时的归一化均方误差为 1.9%，而多测量稀疏贝叶斯算法仅需 15% 采样率即
可达到相似效果且耗时更少。
关键词：压缩感知；单测量向量模型；多测量向量模型；全聚焦方法；全矩阵捕捉
中图法分类号: TB553 文献标识码: A 文章编号: 1000-310X(2023)03-0523-06
DOI: 10.11684/j.issn.1000-310X.2023.03.010

Joint sparse reconstruction application of ultrasonic full matrix data using

multi-measurement vectors model

YAN Fan LIN Li JIN Shijie

(NDT & E Laboratory, Dalian University of Technology, Dalian 116085, China)

Abstract: To overcome the low reconstruction accuracy and long reconstruction time for processing ultrasonic
full matrix capture data by conventional compressed sensing method, e.g., single measurement vector (SMV)
model, this paper studies the feasibility of multiple measurement vectors model (MMV) model. Focusing on
the aluminum alloy specimen with Φ2 mm side-drilled holes in different depths, the multiple sparse Bayesian
learning (MSBL) algorithm in the MMV model and the sparse Bayesian learning (SBL) algorithm in the SMV
model were used to reconstruct ultrasound full matrix data and perform total focusing method, respectively.
Then, normalized root mean square error and array performance factor were introduced to evaluate the recon-
struction effect of images and signals. The experimental results show that the normalized root mean square
error for SBL algorithm is 1.9% for 25% sampling rate and the MSBL algorithm only employs 15% sampling
rate to achieve similar results with less cost time.
Keywords: Compressed sensing; Single measurement vector; Multiple measurement vectors; Total focusing
method; Full matrix capture

2022-02-19 收稿; 2022-04-12 定稿
辽宁省 “兴辽英才计划” 项目 (XLYC1902082), 国家自然科学基金项目 (51905079)
∗
作者简介: 严凡 (1997– ), 男, 江西南昌人, 硕士研究生, 研究方向: 相控阵超声及成像检测。
† 通信作者 E-mail: jinshijie@dlut.edu.cn