水声目标探测和识别融合技术发展综述

A Review of Underwater Acoustic Target Detection and Recognition Technology Based on Information Fusion

水下空间的主动权竞争越来越激烈,针对复杂的海洋环境和多样的目标类型,现有水面水下单平台单传感器探测能力有限、稳定性差,很难满足未来对水声目标的准确探测和识别需求。随着水下探测手段增加以及多源信息融合技术的迅速发展,研究基于融合理论的水声目标探测与识别技术,可实现对水声目标更准确的状态估计和属性判别,进而提高水下探测预警能力。本文首先对现有水面水下探测传感器和平台进行了简要介绍,在此基础上,讨论了水声目标立体观测系统的网络组成和结构特点,并给出一种跨域异构水声目标立体探测系统设计设想,根据现有融合结构给出了基于多信息源的水声目标融合功能框架;然后根据水声融合探测技术的具体应用,分别论述和分析了基于目标信号级和非目标信号级融合检测方法,基于时空特征、运动特征和海洋物理场的融合定位方法,以及基于经典非线性滤波、数据关联和随机集理论的水声目标跟踪方法;最后重点总结了水下目标的各物理场特征,并针对水声识别中普遍存在的目标信息不确定性问题,归纳了基于Bayes理论、D-S证据理论和深度学习算法的三类典型的融合识别方法及框架。指出了基于信息融合的水声目标探测和识别技术所具有的优势和前景,并分析了水下立体观测网络、融合结构、融合探测和识别算法以及工程应用等方面面临的问题和挑战。

The competition for initiative in underwater space was becoming more and more fierce.In view of the complex Ma-rine environment and diverse target types,the detection capability of the existing underwater platform and single sensor was limited and the stability was poor,which was difficult to meet the needs of accurate detection and identification of underwater acoustic targets in the future.With the increase of underwater target detection methods and means and the rapid development of multi-source information fusion technology,the research on underwater acoustic target detection and recognition technology based on information fusion theory can achieve more accurate state estimation and attribute discrimination of underwater acoustic targets,and thus improved the underwater detection and early warning ability.Firstly,the existing detection sensors and detection platforms above and below water were briefly introduced.On this basis,the network composition and structure characteristics of the stereo observation system of underwater acoustic target were discussed,and a cross-domain heteroge-neous stereo observation system design was proposed.According to the existing fusion structure,the functional framework of the multi-information source based underwater acoustic target observation system was given.Then,according to the specific application of underwater acoustic target fusion detection technology,the fusion detection method based on target signal level and non-target signal level,the fusion location method based on space-time characteristics,motion characteristics and Ma-rine physical field,and the underwater acoustic target tracking method based on classical nonlinear filtering,data association and random set theory were discussed and analyzed respectively.The problems faced by each method were analyzed briefly.Finally,the paper focused on the summary and analysis of the physical field characteristics of underwater acoustic targets.In view of the uncertainty of target information,three typical fusion recognition methods and frameworks based on Bayes theory,D-S evidence theory and deep learning algorithm were summarized,and their development status was analyzed.The advan-tages and prospects of underwater acoustic target detection and recognition technology based on information fusion are pointed out,and the problems and challenges in underwater stereo observation network,fusion structure,fusion detection and recog-nition algorithm and engineering applications are analyzed.