Page 194 - 《应用声学》2025年第2期

第 44 卷 第 2 期 Vol. 44, No. 2
2025 年 3 月 Journal of Applied Acoustics March, 2025

⋄ 研究论文 ⋄

声学图像的水下小目标三维形状恢复

易 兵 1,2 蒋立军 1† 刘 佳 1 许 枫 1

(1 中国科学院声学研究所 海洋声学技术实验室 北京 100190)

(2 中国科学院大学 北京 100049)
摘要：基于二维声呐图像的三维场景恢复在水下救援等应用场景中具有重要意义。单一的三维重构算法难以
同时实现图像细节信息的恢复与背景噪声的抑制，使得目标恢复精度较低。针对这一问题，提出了一种数据融
合的三维重构方法。该方法首先利用 C 均值模糊聚类算法提取图像的目标区域与背景区域，然后针对目标区
域采用细节恢复较好的三维重构算法，背景区域选取噪声抑制能力较强的恢复算法，并将二者恢复结果进行
融合，从而得到图像细节信息恢复较好及抑制背景噪声的能力较强的重构结果，提升三维恢复结果的精度。实
验数据处理表明，该方法与传统方法相比，其重构结果的整体平均误差从 24.3% 下降至 10%，说明该方法能更
精准地重构声呐图像的三维形状。
关键词：声呐图像；三维重构；图像分割；数据融合
中图法分类号: O427.9 文献标识码: A 文章编号: 1000-310X(2025)02-0454-09
DOI: 10.11684/j.issn.1000-310X.2025.02.020

Three-dimensional shape restoration of small underwater targets based on

acoustic images

1
YI Bing 1,2 , JIANG Lijun , LIU Jia and XU Feng 1
1
(1 Laboratory of Marine Acoustic Technology, Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China)
(2 University of Chinese Academy of Sciences, Beijing 100049, China)

Abstract: Three-dimensional scene restoration based on two-dimensional sonar images is of great significance
in application scenarios such as underwater rescue. Because it is difficult for a single three-dimensional re-
construction algorithm to realize the recovery of image detail information and the suppression of background
noise at the same time, the target recovery accuracy is low. In order to solve this problem, a three-dimensional
reconstruction method based on data fusion was proposed. In this method, the C-means fuzzy clustering algo-
rithm is used to extract the target area and the background area of the image, and then the three-dimensional
reconstruction algorithm with better detail recovery is used for the target area, and the recovery algorithm
with strong noise suppression ability is selected for the background area, and the recovery results of the two
are fused, so as to obtain the better recovery of image detail information and the stronger reconstruction
results of suppressing background noise, and improve the accuracy of the three-dimensional recovery results.
Experimental data processing shows that compared with the traditional method, the overall average error of
the reconstruction results of the proposed method decreases from 24.3% to 10%, indicating that the proposed
method can reconstruct the three-dimensional shape of sonar images more accurately.
Keywords: Sonar images; Three-dimensional reconstruction; Image segmentation; Data fusion

2023-10-13 收稿; 2023-12-06 定稿
中国科学院青年创新促进会项目 (2020023), 海南省重大科技计划项目 (ZDKJ2020010)
∗
作者简介: 易兵 (1998– ), 男, 四川成都人, 硕士研究生, 研究方向: 信号与信息处理。
† 通信作者 E-mail: jlj@mail.ioa.ac.cn