首页 | 本学科首页   官方微博 | 高级检索  
相似文献
 共查询到17条相似文献,搜索用时 265 毫秒
1.
对噪声和异常值较敏感、鲁棒性差是超限学习机(ELM)的主要问题.在1-范数损失函数的基础上,提出截断1-范数损失函数来抑制噪声和异常值的影响,建立了基于截断1-范数损失函数的鲁棒ELM模型.通过迭代重赋权算法求解对应的优化问题,并利用4个模拟数据集和9个真实数据集验证模型的有效性.数值实验结果表明,在噪声环境下鲁棒ELM的泛化性能优于对比方法,并且具有较强的鲁棒性,尤其是在异常值比例较大的情形下.  相似文献   

2.
研究了一类不确定时滞非线性系统的鲁棒镇定问题.不确定参数是时变未知有界的,但不一定满足所谓的匹配条件.利用线性矩阵不等式,一些充分条件被获得,保证了相应闭环系统的鲁棒稳定性,并在此基础上得出了相应的状态反馈控制器.最后,通过一个数值例子说明了该方法的有效性.  相似文献   

3.
针对空间异构环境下易产生非线性和不确定性影响,多个异构航天器进行姿态控制存在执行器输入饱和约束,对有限的资源进行协作调度难度大等问题,提出了异构航天器包含饱和输入和不确定性动态的鲁棒一致性追踪控制策略.在该控制策略中,参考系统的动态权重矩阵被允许是完全未知的,因此现有的方法并没有可行解.通过引入有向生成树假设,一种新的权重平均方法被提出来构造分布式观测器动态.通过综合分布式观测和非线性动态,一类新的鲁棒一致性追踪控制器被设计.仿真结果表明:当同步行为和分布式观测动态能够被同时获得并且达到领导者状态,异构系统的鲁棒一致性问题将被解决.同时,通过利用多个航天器的动态模型,设计了不同性能指标的参数条件,验证和分析了提出方法的有效性.  相似文献   

4.
面向食品领域的图像检索和分类等方面的研究成为多媒体分析和应用领域越来越受关注的研究课题之一.当前的主要研究方法基于全图提取视觉特征,但由于食品图像背景噪音的存在使得提取的视觉特征不够鲁棒,进而影响食品图像检索和分类的性能.为此,本文提出了一种基于Faster R-CNN网络的食品图像检索和分类方法.首先通过Faster R-CNN检测图像中的候选食品区域,然后通过卷积神经网络(CNN)方法提取候选区域的视觉特征,避免了噪音的干扰使得提取的视觉特征更具有判别力.此外,选取来自视觉基因库中标注好的食品图像集微调Faster R-CNN网络,以保证Faster R-CNN食品区域检测的准确度.在包括233类菜品和49 168张食品图像的Dish-233数据集上进行实验.全面的实验评估表明:基于Faster R-CNN食品区域检测的视觉特征提取方法可以有效地提高食品图像检索和分类的性能.  相似文献   

5.
在集合卡尔曼滤波方法中,根据预报集合的统计特性提供的预报误差协方差矩阵对资料同化起决定性作用。但协方差矩阵低估会引起资料同化滤波发散问题。通过将集合转换卡尔曼滤波方法和时间局地化的H_∞滤波方法相结合,提出一种基于鲁棒集合滤波思想的资料同化方法,放大转移矩阵的特征值,改善估计效果。主要思路是在集合滤波的框架下,按照鲁棒滤波的最小最大准则,实现同化系统性能的改进。利用非线性Lorenz-96混沌系统,考察集合时间局地化的H_∞滤波在系统参数变化时,对同化系统鲁棒性的影响。结果表明:集合时间局地化的H_∞滤波对系统参数变化具有很好的鲁棒性;与传统的滤波方法相比,鲁棒滤波方法提高了同化的效果。  相似文献   

6.
图像去噪是对图像进行进一步处理的基础,结合椒盐噪声信息与传统选择式掩模法提出了一种新的有效方法.该方法根据椒盐噪声的特点,采用噪声检测方法检测出可能是噪声的像素,对这些像素采用加权的选择式掩模法进行处理,而对非疑似噪声像素不做任何处理;在去除图像噪声的同时,可保持图像的细节和清晰度;对去除椒盐噪声效果较好,并且容易实现.给出了不同噪声量的情况下试验的对比图以及不同情况下的峰值信噪比的统计表.实验证明,该方法 能得到很好的去噪效果.  相似文献   

7.
图像分割是计算机视觉中基础且重要的一个问题.熵阈值图像分割作为一种有效的分割方法,被广泛应用于模式识别和图像处理中.传统的图像分割方法并不能获得足够有效的图像特征.为解决这个问题且进一步探究熵阈值在图像分割中的应用,引入一种GLLE(Gray Level and Local Entropy)二维直方图改进熵阈值图像分割模型,并提出了基于模糊熵的方法计算所建立的二维直方图模型.通过标准实验数据集上的对比实验表明,基于模糊熵的GLLE熵阈值分割方法可以得到更加准确的阈值,提高了分割精度.同时在处理不同类型图像的表现上优于往常的算法,具有更强的鲁棒性.  相似文献   

8.
分别针对具有结构参数和范数有界参数的滞后型Lurie控制系统的鲁棒绝对稳定性问题,利用Lyapunov方法给出了系统鲁棒绝对稳定的时滞无关条件及时滞相关条件.得到的结果用线性矩阵不等式(LMI)表示,易于利用MATLAB工具箱求得保守性较低的条件.  相似文献   

9.
通过类Lyapunov方法,结合线性矩阵不等式(LMI)技术,给出了具有时滞的Lipschitz非线性离散系统全维观测器存在的充分条件.在此基础上,设计了具有扰动的时滞Lipschitz非线性离散系统的鲁棒观测器,提出了该鲁棒观测器存在的一个充分条件,并用具体例子说明了其可行性与有效性.  相似文献   

10.
研究了一类时变不确定系数多离散时滞中立型系统的时滞相关鲁棒稳定性问题.利用线性矩阵不等式技术,提出了一种计算该系统自由权矩阵和时滞上界的简化方法,并在系统中考虑了中立项时滞和离散时滞,得到了上述系统为稳定与鲁棒稳定的一些充分条件.最后,用数值例子说明了该方法的有效性和较小的保守性.  相似文献   

11.
In mountain areas, radar observations are often contaminated (1) by echoes from high-speed moving vehicles and (2) by point-wise ground clutter under either normal propagation (NP) or anomalous propagation (AP) conditions. Level Ⅱ data are collected from KMTX (Salt Lake City, Utah) radar to analyze these two types of contamination in the mountain area around the Great Salt Lake. Human experts provide the "ground truth" for possible contamination of either type on each individual pixel. Common features are then extracted for contaminated pixels of each type. For example, pixels contaminated by echoes from high-speed moving vehicles are characterized by large radial velocity and spectrum width. Echoes from a moving train tend to have larger velocity and reflectivity but smaller spectrum width than those from moving vehicles on highways. These contaminated pixels are only seen in areas of large terrain gradient (in the radial direction along the radar beam). The same is true for the second type of contamination - pointwise ground clutters. Six quality control (QC) parameters are selected to quantify the extracted features. Histograms are computed for each QC parameter and grouped for contaminated pixels of each type and also for non-contaminated pixels. Based on the computed histograms, a fuzzy logical algorithm is developed for automated detection of contaminated pixels. The algorithm is tested with KMTX radar data under different (clear and rainy) weather conditions.  相似文献   

12.
轨道误差和长波大气延迟组成的系统误差是影响InSAR形变监测精度的重要因素之一.传统方法在空间域对干涉图的系统误差建模,容易导致长波形变和系统误差相混淆.本文在时空域利用附加系统参数对系统误差建模,同时根据观测值质量对差分相位观测值定权,采用附加系统参数的加权最小二乘法估计形变参数和系统误差,实现了长波形变和系统误差的分离.模拟实验结果表明,在形变与系统误差的空间变化特性完全一致的极端情况下,本文方法能实现两者的有效分离,估计的形变速率均方根误差比传统方法降低了98.8%.ASAR数据实验显示当形变尺度较小且分散分布时,本文方法和传统方法得到的结果相似;当形变在研究区内表现为长波变化时,本文方法比传统方法估计的形变结果更为稳健.  相似文献   

13.
14.
An improved velocity volume processing method   总被引:3,自引:0,他引:3  
Velocity volume processing (VVP) retrieval of single Doppler radar is an effective method which can be used to obtain many wind parameters. However, due to the problem of an ill-conditioned matrix arising from the coefficients of equations not being easily resolved, the VVP method has not been applied adequately and effectively in operation. In this paper, an improved scheme, SVVP (step velocity volume processing), based on the original method, is proposed. The improved algorithm retrieves each group of components of the wind field through a stepwise procedure, which overcomes the problem of an ill-conditioned matrix, which currently limits the application of the VVP method. Variables in a six-parameter model can be retrieved even if the analysis volume is very small. In addition, the source and order of errors which exist in the traditional method are analyzed. The improved method is applied to real cases, which show that it is robust and has the capability to obtain the wind field structure of the local convective system. It is very helpful for studying severe storms.  相似文献   

15.
对Gong(2003)提出的基于Modified VAD技术的多普勒雷达径向速度退模糊方法进行了改进。原方法从各高度上径向风速的方位梯度信息中估计水平风廓线的分布,在使用各高度参考风进行一次退模糊的基础上,进行连续性检验完成整个退模糊过程。使用参考风进行退模糊能在很大程度上提高退模糊的效率。但是由于中小尺度系统的扰动和复杂的采样覆盖的影响,一般很难获得比较准确的估计水平风速,这将对进一步的退模糊过程产生很大的影响。针对这个问题,在水平风估计过程增加了质量控制参数,使其能够更好地估计水平风。此外,还对原方法中的几项参数的选择进行了讨论和调整。通过对两个个例的退模糊试验,检验了改进后方法的效果。  相似文献   

16.
In this paper, we propose an improved multisite weather generation with applications to the historical data in South Korea. The proposed method improve the algorithm of Wilks (1998, 1999) by automatically selecting an optimal model that represents precipitation amounts and by providing a procedure to obtain a symmetric positive definite estimate for the covariance matrix. The proposed method is computationally fast, and hence, it can be feasible to handle a massive data. We apply the proposed method to the precipitation and temperature data collected 170 stations in South Korea for the period 1976–2005 which are given by the Korea Meteorological Administration (KMA). Results of the proposed method demonstrate the promising performance in terms of spatial correlation and long-term variation as compared with those of the multisite method of Wilks (1998) and the single-site weather generator.  相似文献   

17.
ICSED (Improved Cluster Shade Edge Detection) algorithm and other various methods to accurately and efficiently detect edges on satellite data are presented. Error rate criterion is used to statistically evaluate the perform-ances of these methods in detecting oceanic features for both noise free and noise contaminated AVHRR (Advanced Very High Resolution Radiometer) IR image with Kuroshio. Also, practical experiments in detecting the eddy of Kuroshio with these methods are carried out for comparison. Results show that the ICSED algorithm has more advantages than other methods in detecting mesoscale features of ocean. Finally, the effectiveness of window size of ICSED method to oceanic features detection is quantitatively discussed.  相似文献   

设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号