Page 148 - 《应用声学》2022年第3期

470 2022 年 5 月

MobilenetV2 网络与原网络相比，参数量及计算资 [10] Sinai A, Amar A, Gilboa G. Mine-like objects detectionin
源消耗减少，分类准确率进一步提高，在水下目标分 side-scan sonar images using a shadows-highlights geomet-
rical features space[C]. OCEANS 2016 MTS/IEEE Mon-
类任务中具有更好的性能。相比融合前的网络，融
terey, Monterey, CA, 2016: 1–6.
合网络模型的学习曲线收敛更快，且准确率更高，在 [11] Williams D P. Underwater target classification in syn-
add、maximum、concatenate三种融合算法下，融合 thetic aperture sonar imagery using deep convolutional
网络的分类准确率均有不同程度的提升。在单路数 neural networks[C]. 2016 23rd International Confer-
ence on Pattern Recognition (ICPR), Cancun, 2016:
据缺失的情况下，融合网络的分类准确率仍能达到 2497–2502.
85%以上，具有一定的鲁棒性。 [12] 朱可卿, 田杰, 黄海宁. 水下高分辨率声图中小目标的深度网
络分类方法 [J]. 声学学报, 2019, 44(4): 595–603.
Zhu Keqing, Tian Jie, Huang Haining. Underwater ob-
参 考 文 献 jects classification method in high-resolution sonar images
using deep neural network[J]. Acta Acustica, 2019, 44(4):
[1] 林森, 赵颍. 水下光学图像中目标探测关键技术研究综述 [J]. 595–603.
激光与光电子学进展, 2020, 57(6): 26–37. [13] Howard A G, Zhu M, Chen B, et al. MobileNets: efficient
Lin Sen, Zhao Ying. Review on key technologies of tar- convolutional neural networks for mobile vision applica-
get exploration in underwater optical images[J]. Laser & tions[J]. IEEE Access, 2017, 6: 1–14.
Optoelectronics Progress, 2020, 57(6): 26–37. [14] 孟琭, 徐磊, 郭嘉阳. 一种基于改进的 MobileNetV2 网络语义
[2] Gleason A C R, Shihavuddin A, Gracias N, et al. Im- 分割算法 [J]. 电子学报, 2020, 48(9): 1769–1776.
proved supervised classification of underwater military Meng Lu, Xu Lei, Guo Jiayang. Semantic segmentation
munitions using height features derived from optical im- algorithm based on improved MobileNetV2[J]. Acta Elec-
agery[C]. OCEANS 2015-MTS/IEEE Washington, Wash- tronica Sinica, 2020, 48(9): 1769–1776.
ington, DC, USA, 2015: 1–5. [15] Sandler M, Howard A, Zhu M, et al. Mo-
[3] Pramunendar R A, Wibirama S, Santosa P I. Fish clas- bileNetV2: inverted residuals and linear bottlenecks[C].
sification based on underwater image interpolation and 2018 IEEE/CVF Conference on Computer Vision and
back-propagation neural network[C]. 2019 5th Interna- Pattern Recognition, Salt Lake City, UT, USA, 2018:
tional Conference on Science and Technology (ICST), Yo- 4510–4520.
gyakarta, Indonesia, 2019: 1–6. [16] Bousbai K, Merah M. A comparative study of hand ges-
[4] 王士龙, 徐玉如, 万磊, 等. 基于边界矩和改进 FCM 聚类 tures recognition based on MobileNetV2 and ConvNet
的水下目标识别 [J]. 系统工程理论与实践, 2012, 32(12): models[C]. 2019 6th International Conference on Image
2809–2815. and Signal Processing and their Applications (ISPA),
Wang Shilong, Xu Yuru, Wan Lei, et al. Underwater tar- Mostaganem, Algeria, 2019: 1–6
gets recognition based on contour moment and modified [17] Thakkar V, Tewary S, Chakraborty C. Batch normal-
FCM algorithm[J]. System Engineering-Theory & Prac- ization in convolutional neural networks—A compara-
tice, 2012, 32(12): 2809–2815. tive study with CIFAR-10 data[C]. 2018 Fifth Interna-
[5] 孙宝申. 高分辨侧扫及合成孔径声纳图像与回波中类似水雷 tional Conference on Emerging Applications of Informa-
目标计算机检测与分类 [J]. 应用声学, 2013, 32(4): 305–311. tion Technology (EAIT), Kolkata, India, 2018: 1–5.
Sun Baoshen. Computer aided detection and classifica- [18] Hoang V, Hoang V, Jo K. Realtime multi-person pose
tion of mine-like objects in high-resolution side scan and estimation with RCNN and depthwise separable convo-
synthetic aperture sonar images and waveforms[J]. Jour- lution[C]. 2020 RIVF International Conference on Com-
nal of Applied Acoustics, 2013, 32(4): 305–311. puting and Communication Technologies (RIVF), Ho Chi
[6] 刘纪元. 合成孔径声呐技术研究进展 [J]. 中国科学院院刊, Minh City, Vietnam, 2020: 1–5.
2019, 34(3): 283–288. [19] 康硕, 柯臻铮, 王璇, 等. 基于红外和可见光图像融合的铺丝
[7] Raj M V, Murugan S S. Underwater image classification 缺陷检测方法 [J]. 航空学报, 2022, 43(3): 425187.
using machine learning technique[C]. 2019 International Kang Shuo, Ke Zhenzheng, Wang Xuan, et al. Detection
Symposium on Ocean Technology (SYMPOL), Ernaku- method of defects in automatic fiber placement based on
lam, India, 2019: 166–173. infrared and visible image fusion[J]. Acta Aeronautica et
[8] Sperle M, Negri E, Ternes C. Automatic classification Astronautica Sinica, 2022, 43(3): 425187.
of sidescan sonar images for mapping marine mineral [20] Wolpert A, Teutsch M, Sarfraz M S, et al. Anchor-free
resources[C]. 2015 IEEE/OES Acoustics in Underwater small-scale multispectral pedestrian detection[J]. arXiv:
Geosciences Symposium (RIO Acoustics), Rio de Janeiro, 2008.08418.
Brazil, 2015: 1–5. [21] 刘天宝, 吴晓潭, 黄勇, 等. 三维成像声呐图像重建研究 [J]. 声
[9] Cho H, Gu J, Yu S. Robust sonar-based underwater ob- 学技术, 2015, 34(4): 358–361.
ject recognition against angle-of-view variation[J]. IEEE Liu Tianbao, Wu Xiaotan, Huang Yong, et al. Image re-
Sensors Journal, 2016, 16(4): 1013–1025. construction for 3D acoustical imaging sonar[J]. Technical
Acoustics, 2015, 34(4): 358–361.