基于混沌特性与随机共振特性表征的水中弱小目标信息融合探测

According to the demand of underwater dim and small targets detection aiming at improving utilization and development of ocean resources, safeguarding marine rights and interests, and securing marine national defense, we have proposed to investigate: 1) detection and representing method of chaotic characteristics of underwater dim and small target based on chaos theory, which could represent four different kinds of chaotic characteristic quantities, including Lyapunov exponent, association dimensions, responding characteristics of chaotic information and system complexity, of ocean ambient noises and radiated noises of underwater targets, with emphasis on the study of the difference among chaos characteristics of different underwater targets; 2) detection and method of enhancing line spectrum of underwater dim and small targets based on stochastic resonance. We will construct an stochastic resonance system of underwater dim and small target signals which can be utilized to represent both characteristic line spectrum and instantaneous one. The emphasis is on stochastic resonance method with large parameters and weak signal detection based on Duffing oscillators under the condition of low signal-to-noise ratio (SNR); 3) the way to break through the bottleneck of detection difficulties on underwater dim and small targets under low SNR condition; 4) information fusion method for target detection based on kernel function-based KPCA and SVM. We will design kernel functions and KPCA, by which a characteristic information fusion method based on SVM with highlighted performance could be proposed. Improvement of effectiveness and reliability of underwater dim and small target detection could then be expected. The theory and method of proposed underwater dim and small target detection could be further verified and perfected by conducting experiments. We expect our research to lay the foundation for accelerating the utilization and development of ocean resources and marine national defense security.

针对海洋利用开发、海洋权益维护与安全等对水中弱小目标探测的需求，研究基于混沌理论的水中弱小目标混沌特性表征方法与探测，表征海洋环境噪声和水中目标辐射噪声混沌特性的Lyapunov指数、关联维数、混沌信息通过特性和系统复杂度4种特征量，重点研究用于探测的不同水中目标混沌特征之间的差异性；研究基于随机共振的水中弱小目标线谱特征增强方法与探测，构建水中弱小目标信号的随机共振系统，表征特征线谱、瞬时线谱，重点研究低信噪比条件下，混沌duffing振子的大参数随机共振方法与微弱信号探测；突破水中弱小目标在低信噪比条件下探测难的瓶颈。研究基于核函数技术的KPCA和 SVM的信息融合探测方法，设计核函数和KPCA，构建高性能的支持向量机特征信息融合方法，提高对水中弱小目标探测的有效性和可靠性。通过试验，进一步验证并完善水中弱小目标探测理论方法。为推进我国海洋的利用开发和海军国防安全提供理论支持。

本项目围绕热点的非线性混沌理论以及随机共振理论，通过混沌特性和随机共振特性表征微弱信号，系统而深入地研究了非线性微弱信号检测理论及应用方法，完成了全部研究内容，突破了复杂环境下水中弱小目标探测难的瓶颈。1、建立了多维非线性混沌量化表征分析模型，通过理论及实船分析验证的方式给出了目标非线性特性的基本规律；2、围绕目标信号中特殊弱混沌特性的特殊相空间识别，提出了一种Duffing振子的相空间图形离散分布列计算方法，通过类网格函数实现了对系统输出的嵌入式表征，在无先验信息时可以检测到更低信噪比下的目标；3、针对低信噪比下输入噪声对系统的扰动，提出了一种基于临界阈值字典的双Duffing系统逆向相变方法，解决了目标信号相位和内策动力相位不同步问题，并推导了相位的不遍历性导致的虚惊概率，大幅度降低了最低检测信噪比；4、研究了双稳态系统响应的基本规律，理论推导了随机共振的数学匹配关系，建立了信噪比增益为核心的目标函数匹配优化数学模型；5、理论分析了经典过阻尼系统在求解非线性响应时忽略阻尼的问题，建立了一般化的阻尼双稳态系统，提出了一阶阻尼匹配随机共振的微弱信号检测方法，提升了低信噪比条件下线谱被动检测的性能；6、考虑实际复杂背景噪声环境，建立了列维分布噪声下的二阶双稳态随机共振系统，提升了在强冲击背景噪声下的检测性能；7、研究了随机共振输出响应的多测度表征方法，提出了多测度融合随机共振的特征检测与提取方法；8、提出了一种自适应匹配随机共振的微弱时变特征线谱增强方法，能够有效增强目标特征线谱，并验证了该方法的实用性能；9、提出了基于贝叶斯分类器的Wrapper 算法，通过设计分类正确率的置信区间上、下界指标，提升了特征优化的稳定性；10、提出基于最大信息系数的水下目标的特征融合分类方法，构建目标函数对特征子集的相关度、冗余度进行评估，实现最优特征融合；11、考虑复杂环境下目标时变特征提取难的问题，提出了基于结构化张量分解模型实现水声特征融合，通过对多种特征的高维张量处理，提高分类算法在低信噪比环境中的鲁棒性。本项目的研究，完善了水中弱小目标微弱信息检测与处理的理论方法体系，为推进我国海洋的利用开发和海军国防安全有着重要的理论意义和应用价值。