随着海洋勘探技术的发展，水声图像在海洋开发中占据越来越重要的地位。针对前视声呐在障碍物识别中存在的高噪声以及目标成像不连续等问题，借鉴神经网络中卷积核的思想，将形态学算子视为特殊的卷积核。同时，将帧间差分图像与当前帧所获取图像的特征相融合，提出一种基于特征融合的目标检测算法。而后利用Mean-shift(均值漂移)算法对检测后得到的清晰的障碍物目标进行跟踪。实验结果表明，该算法能够有效地抑制动态背景中的噪声，同时使障碍物目标能够以连续、稳定变化的运动形态，清晰地呈现在连续多帧图像中，并对目标进行有效地跟踪。以“探索1000”AUV在千岛湖采集的声呐视频序列作为实验数据，表明该算法具有广阔的应用前景。
With the development of ocean exploration technology, underwater acoustic images occupy an increasingly important position in ocean development. Aiming at the problems of high noise and discontinuous target imaging in the obstacle recognition of forward-looking sonar, the idea of convolution kernel in neural network is used for reference, and the morphological operator is regarded as a special convolution kernel, At the same time, the inter-frame difference image is fused with the features of the image acquired in the current frame, and a target detection algorithm based on feature fusion is proposed. Then, the Mean-shift algorithm is used to track the clear obstacle targets obtained after detection. The experimental results show that the algorithm can effectively suppress the noise in the dynamic background, at the same time, enable the obstacle target to be clearly presented in continuous multi-frame images in a continuous and stable changing motion form, and to effectively track the target. The sonar video sequence collected by "Exploration1000" AUV in Qiandao Lake is used as experimental data, which shows that this algorithm has broad application prospects.
2021,43(11): 143-148 收稿日期：2020-12-01
DOI：10.3404/j.issn.1672-7649.2021.11.027
分类号：TP391;TN911.73
基金项目：国家重点研发计划（2017YFC0305703）
作者简介：刘昊搏(1997-)，男，硕士研究生，研究方向为声呐图像处理与目标检测
参考文献：
[1] 方世良, 杜栓平, 罗昕炜, 等. 水声目标特征分析与识别技术[J]. 中国科学院院刊, 2019, 34(3): 297–305
FANG Shi Liang, DU Shuan Ping, LUO Xin Wei, et al. Feature analysis and recognition technology of underwater acoustic target[J]. Bulletin of Chinese Academy of Sciences, 2019, 34(3): 297–305
[2] 孙亚晶. 基于深度学习的声呐图像目标识别技术研究[D]. 南京: 东南大学, 2019.
SUN Ya Jing. Research of Sonar Image Target Recognition Technology Based On Deep Learning[D]. Nanjing: Southeast University, 2019.
[3] 李庆武, 霍冠英, 周妍. 声呐图像处理[M]. 北京: 科学出版社, 2015: 18−22.
[4] 温永媛. 合成孔径声呐成像旁瓣抑制算法研究[D]. 昆明: 云南大学, 2016.
WEN Yong Yuan. Sidelobe suppression algorithm of synthetic aperture sonar imaging[D]. Kunming: Yunnan University, 2016.
[5] 徐从营, 裴承艳. 基于分区自适应阈值的声呐障碍物检测算法研究[J]. 数字海洋与水下攻防, 2018, 1(2): 1–5
XU Cong Ying, FEI Cheng Yan. Research on obstacle detection algorithm of sonar based on partition adaptive threshold[J]. Digital Ocean & Underwater Warfare, 2018, 1(2): 1–5
[6] 屈晶晶, 辛云宏. 连续帧间差分与背景差分相融合的运动目标检测方法[J]. 光子学报, 2014, 43(7): 219–226
QU Jing Jing, XIN Yun Hong. Combined continuous frame difference with background difference method for moving object detection[J]. Acta Photonica Sinica, 2014, 43(7): 219–226
[7] 郭戈, 王兴凯, 徐慧朴. 基于声呐图像的水下目标检测、识别与跟踪研究综述[J]. 控制与决策, 2018, 33(5): 906–922
GUO Ge, WANG Xing Kai, XU Hui Pu. Survey of research on underwater target detection, recognition and tracking based on sonar images[J]. Journal of Control and Decision, 2018, 33(5): 906–922
[8] S. Yu, T. Kim, G. Marani and S. K. Choi, Real-Time 3D sonar image recognition for underwater vehicles[C]//2007 Symposium on Underwater Technology and Workshop on Scientific Use of Submarine Cables and Related Technologies, Tokyo, 2007, 142−146, doi: 10.1109/UT.2007.370843.
[9] 崔杰, 胡长青, 徐海东. 基于帧差法的多波束前视声呐运动目标检测[J]. 仪器仪表学报, 2018, 39(2): 169–176
CUI Jie, HU Chang Qing, XU Hai Dong. Moving target detection for multi-beam forward-looking soanr based on frame-difference method[J]. Chinese Journal of Scientific Instrument, 2018, 39(2): 169–176
[10] 张俞鹏, 刘志, 任静茹, 等. 基于多波束前视声呐的水下障碍物检测及避障算法[J]. 工业控制计算机, 2020, 33(3): 6–8
ZHANG Yu Peng, Liu Zhi, Ren Jing Ru, et al. Underwater obstacle detection and avoidance based on multi-beam forward-looking sonar[J]. Industrial Control Computer, 2020, 33(3): 6–8
[11] 陈双叶, 王善喜. 基于五帧差分和改进的Meanshift算法的运动目标跟踪[J]. 计算机科学, 2016(S1): 203–206
CHEN Shuang Ye, WANG Shan Xi. Moving object tracking based on five frame difference and improved meanshift algorithm[J]. Computer Science, 2016(S1): 203–206
[12] J. K. Suhr, H. G. Jung, G. Li and J. Kim, Mixture of Gaussians-Based Background Subtraction for Bayer-Pattern Image Sequences[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2010, 21(3): 365−370.
[13] 张少娜. 基于均值漂移的目标跟踪算法研究[D]. 南京: 南京邮电大学, 2013.
ZHANG Shao Na. Research on mean shift based object tracking algorithm[D]. Nanjing: Nanjing University of Posts and Telecommunications, 2013.
