Page 33 - 《应用声学》2021年第4期

第 40 卷 第 4 期 逄岩等： 基于 Gammatone 滤波器组时频谱和卷积神经网络的海底底质分类 517

ral networks[J]. Journal of Chinese Mini-Micro Computer
4 结论 Systems, 2019, 40(9): 1825–1831.
[9] Parkhi O M, Vedaldi A, Zisserman A. Deep face recogni-
本文将深度学习的思想应用到海底底质分类 tion[C]// British Machine Vision Conference, 2015.
当中，将底质图像信号的 Gammatone 滤波器组时 [10] 闫琰. 基于深度学习的文本表示与分类方法研究 [D]. 北京:
北京科技大学, 2016.
频谱作为 CNN 模型的输入，进行底质的分类识别， [11] Berthold T, Leichter A, Rosenhahn B, et al. Seabed sedi-
取得了较高的分类准确率，分类准确率均优于其他 ment classification of side-scan sonar data using convolu-
tional neural networks[C]// 2017 IEEE Symposium Series
常用底质分类的方法。同时，本文进一步验证了所
on Computational Intelligence (SSCI). IEEE, 2017.
提出的方法对于泥底质具有很好的泛化能力，但对 [12] Luo X, Qin X, Wu Z, et al. Sediment classification of
于沙底质和泥底质，泛化能力不强，需要通过增加 small-size seabed acoustic images using convolutional neu-
ral networks[J]. IEEE Access, 2019, PP(99): 1.
CNN 模型层数或者增加沙和石底质训练样本数量
[13] Darling A M. Properties and implementation of the gam-
来进一步完善本文提出的方法。此外，本文采用的 matone filter: a tutorial[J]. Speech Hearing and Language,
底质信号数据集均提取于底质的侧扫声呐图像，未 Work in Progress, 1991: 43–61.
[14] 钱思冲, 向阳, 肖小勇, 等. 基于 Gammatone 滤波器组的内燃
来有待进一步使用原始的底质侧扫数据进行研究
机气缸盖振动特性研究 [J]. 内燃机工程, 2013, 34(6): 36–42.
论证，以实现在实际中的应用。 Qian Sichong, Xiang Yang, Xiao Xiaoyong, et al. Applica-
tion of Gammatone filter bank to vibration characteristics
analysis of engine cylinder head[J]. Chinese Internal Com-
参 考 文 献
bustion Engine Engineering, 2013, 34(6): 36–42.
[15] 刘海燕, 田钢, 石战结. 几种时频分析方法的比较和实际应
[1] 郑红霞, 张训华. 海底底质分类方法综述 [C]// 中国地球物理 用 [J]. CT 理论与应用研究, 2015(2): 199–208.
2013——第二十八专题论文集, 2013. Liu Haiyan, Tian Gang, Shi Zhanjie. The comparison
[2] Blondel P, Sichi O G. Textural analyses of multibeam of time-frequency analysis methods and their applica-
sonar imagery from Stanton Banks, Northern Ireland tions[J]. Computerized Tomography Theory and Appli-
continental shelf[J]. Applied Acoustics, 2008, 70(10): cations, 2015(2): 199–208.
1288–1297. [16] 董建华, 顾汉明, 张星. 几种时频分析方法的比较及应用 [J].
[3] 杨词银, 许枫. 基于二次反锐化掩模的多特征侧扫声纳成像海 工程地球物理学报, 2007, 4(4): 312–316.
底底质分类 [J]. 电子学报, 2005, 33(10): 1841–1844. Dong Jianhua, Gu Hanming, Zhang Xing. A compar-
Yang Ciyin, Xu Feng. Multi-feature seafloor sediments ison of time-frequency analysis methods and their ap-
classification for sidescan sonar imagery based on a plications[J]. Chinese Journal of Engineering Geophysics,
quadratic unsharp masking operator[J]. Chinese Journal 2015(2): 199–208.
of Electronics, 2005, 33(10): 1841–1844. [17] 常亮, 邓小明, 周明全, 等. 图像理解中的卷积神经网络 [J]. 自
[4] Pace N G, Gao H. Swathe seabed classification[J]. IEEE 动化学报, 2016, 42(9): 1300–1312.
Journal of Oceanic Engineering, 1988, 13(2): 83–90. Chang Liang, Deng Xiaoming, Zhou Mingquan, et al.
[5] Tamsett D. Sea-bed characterisation and classification Convolutional neural networks in image understanding[J].
from the power spectra of side-scan sonar data[J]. Marine Acta Automatica Sinica, 2016, 42(9): 1300–1312
Geophysical Researches, 1993, 15(1): 43–64. [18] 倪志伟. BP 网络中激活函数的深入研究 [J]. 安徽大学学报
[6] Chu W, Champagne B. A simplified early auditory model (自然科学版), 1997(3): 48–51.
with application in speech/music classification[C]// Con- Ni Zhiwei. Deep study on activation function in BP net-
ference on Electrical & Computer Engineering. IEEE, work[J]. Journal of Anhui University(Natural Science Edi-
2007. tion), 1997(3): 48–51.
[7] 李允公, 张金萍, 戴丽, 等. 基于听觉模型 ZCPA 的故障 [19] Lecun Y, Bottou L. Gradient-based learning applied to
诊断特征提取方法研究 [J]. 中国机械工程, 2009, 20(24): document recognition[J]. Proceedings of the IEEE, 1998,
2988–2992. 86(11): 2278–2324.
Li Yungong, Zhang Jinping, Dai Li, et al. Study on the [20] van Walree P A, Tęgowski J, Laban C, et al. Acous-
ZCPA-auditory-model-based method of feature extraction tic seafloor discrimination with echo shape parameters: a
for mechanical faults diagnosis[J]. China Mechanical En- comparison with the ground truth[J]. Continental Shelf
gineering, 2009, 20(24): 2988–2992. Research, 2005, 25(18): 2273–2293.
[8] 张泽苗, 霍欢, 赵逢禹. 深层卷积神经网络的目标检测算法综 [21] Atallah L, Smith P J P. Using wavelet analysis to classify
述 [J]. 小型微型计算机系统, 2019, 40(9): 1825–1831. and segment sonar signals scattered from underwater sea
Zhang Zemiao, Huo Huan, Zhao Fengyu. Survey of ob- beds[J]. International Journal of Remote Sensing, 2003,
ject detection algorithm based on deep convolutional neu- 24(21): 4113–4128.