
The Ornstein-Uhlenbeck (OU) process was used to simulate random perturbations in the environment. Considering the influence of telegraph noise and jump noise, a stochastic Gilpin-Ayala nonautonomous competition model driven by the mean-reverting OU process with finite Markov chain and Lévy jumps was established, and the asymptotic behaviors of the stochastic Gilpin-Ayala nonautonomous competition model were studied. First, the existence of the global solution of the stochastic Gilpin-Ayala nonautonomous competition model was proven by the appropriate Lyapunov function. Second, the moment boundedness of the solution of the stochastic Gilpin-Ayala nonautonomous competition model was discussed. Third, the existence of the stationary distribution of the solution of the stochastic Gilpin-Ayala nonautonomous competition model was obtained. Finally, the extinction of the stochastic Gilpin-Ayala nonautonomous competition model was proved. The theoretical results were verified by numerical simulations.
Spectral line enhancement and transient interference suppression of underwater objects are critical issues for passive sonar systems. Conventional approaches for processing spectral lines have focused on either time-domain or frequency-domain methods. In this study, the constrained non-negative matrix factorization is proposed to process the underwater spectral lines in the joint time-frequency domain. Based on the sparsity of spectral lines in the frequency domain, the sparseness criterion is utilized to constrain the basis matrix that represents the frequency-mode of the signal. The correlation between sparsity and frequency estimation accuracy is examined with weight coefficients, and an effective weight coefficient interval for the sparseness term is determined to optimize the detection of the spectral line. To address the issue of abrupt changes in signal energy caused by transient interference, the temporal continuity criterion is applied to constrain the coefficient matrix representing the temporal gain mode of the signal. An analysis is conducted to determine the impact of weight coefficient on the continuity of the coefficient matrix, and the optimal weight coefficient of the temporal continuity term is established to suppress local transient strong interference. Experimental results demonstrate that the algorithm significantly enhances the capacity for the detection and extraction of spectral lines.
The problem of robustness optimization with respect to the gap metric in the context of discrete-time Linear time-varying(LTV)systems is considered,which involves the computation of the maximum stability margin.Based on the inner-outer factorization of operators in nest algebras,a formula for the maximum stability margin is derived.Furthermore,the existence of an optimal controller is proved for uniformly asymptotically time-invariant LTV systems.
无人机(Unmanned aerial vehicle,UAV)遥感图像拼接是指将两幅或多幅具有相似场景内容的高分辨率无人机遥感图像拼接为一幅包含更多信息的大视野图像,在军事和地理测绘等领域得到了广泛应用.传统算法通常依赖于手工特征,无法有效地提取弱纹理图像的特征.若图像之间视差较大时,会导致拼接无法进行.为了解决上述问题,基于计算机视觉组(Visual Geometry Group-16,VGG-16)网络结合孪生网络框架提出了一种用于无人机遥感图像拼接的有监督模型.基于VGG-16 网络设计了权值共享的孪生特征提取网络,解决特征提取不充分的问题.设计了能够回归图像之间空间变换关系的回归网络,并使用分组卷积代替普通卷积以提升网络速度.同时,为了解决将图像之间真实变换关系作为标签的图像拼接数据集难以获取的问题,基于一定程度的仿射变换,构建了自己的数据集.实验结果表明,本方法在无人机遥感图像拼接的主观视觉效果以及客观评价指标上均有较好的结果,与 ORB 算法(Oriented FAST and rotated BRIEF,ORB)和CAU-DHE算法(Content-aware unsupervised deep homography estimation,CAU-DHE)相比,主观视觉上图像拼接精度提升,结构相似性分别提高了约12.4%和2.3%,均方根误差分别降低了约15.0%和4.4%.
基于统计量化分析法对我国唐宋诗词的用词进行研究,综合运用数理统计数据处理方法和其他方法,建立了余弦相似度算法模型和隐含狄利克雷分布(Latent dirichlet allocation,LDA)模型,验证了字词能够很好地代表不同诗词的风格和主题,进一步根据LDA模型分别选取了唐宋时期各3 个主题和30 个关键词,证明主题和关键词对唐宋诗词分类有显著的相关性,并取得了很好的分类效果,最后运用支持向量机(Support vector machines,SVM)算法模型验证了题目、风格和用词等变量对题材分类产生影响的假设,进一步验证了LDA模型结果,也为唐宋诗词的分类和本体研究提供了科学依据.