The stochastic resonance (SR) of an optical bistable system with cross-correlated additive white and multiplicative colored noises and periodic signal is studied using the unified colored noise approximation and the theory of signal-to-noise ratio (SNR). Results show that cross-correlation intensity \lambda enforces the SR of the system. The position of the peak on the SNR-\tau curves moves to the right direction along with the increase of \lambda (\tau is the self-correlation time of the multiplicative colored noise). We find the SR phenomenon in the SNR-D and SNR-Q curves (D and Q are the intensities of the additive and multiplicative noises, respectively), but not in the SNR-\lambda curves.

为了描述局部特征在图像空间中相对位置关系，提出一种局部特征空间相关核(Spatial Correlation Kernel, SCK)用于图像目标分类。该方法首先提取并量化图像中的局部特征，再计算量化后的局部特征的空间位置自相关度，然后利用直方图交叉匹配两幅图像的空间位置自相关度得到局部特征空间相关核。该核充分利用局部特征的强分辨能力及其空间位置，且SCK 具有线性计算复杂度，满足正定条件，可以运用于基于核的学习算法。本文将SCK 嵌入支持向量机对公共数据库中图像目标进行分类，实验结果表明，SCK 可以获得良好的时间效率和分类性能。

本文提出了一种基于标准差特征平面Contrast box滤波的可见光遥感图像舰船目标检测方法。选用局部统计方差作为目标检测特征,实现对不同亮度舰船的统一特征描述,并消除海面平均亮度变化的影响。然后在二维检测特征平面上通过Contrast box滤波自适应确定局部目标检测阈值,并结合目标的空间结构信息完成疑似目标定位。最后借助先验舰船特征模型对疑似目标集合进行验证以去除虚警,输出最终目标检测结果。实验结果表明,该方法对于可见光遥感图像中的舰船目标能够达到99.5%的目标检测准确率,同时目标检测虚警率为5%。

本文通过分析基于小波变换的动态系统模型,提出一种基于小波多尺度的Kalman数据滤波方法,本文利用小波的多尺度特点,把初始估计序列多尺度分解,并在不同尺度层上进行Kalman滤波估计,再利用小波重构来融合各层的估计信息,把标准Kalman滤波只在单一尺度和时间轴上对状态估计值和误差协方差进行数据更新,改进为基于小波变换的尺度轴和时间轴上的双向数据更新,该算法将小波多尺度分解去噪和Kalman滤波相结合,对实际中含较强噪声的动态系统的状态估计效果较好。算法也可用于多分辨率多传感器数据融合。

Considering a single-mode laser system with cross-correlated additive colored noise and multiplicative colored noise, we study the effects of correlation among noises on the normalized intensity correlation function C(s). C(s) is derived by means of the projection operator method. The effects of the self-correlation time \tau1 of the additive colored noise, \tau2 of the multiplicative colored noise, and the effect of the cross-correlation time \tau0 between the two noises on C(s) are discussed by numerical calculation. For the case of positive correlation (\lambda>0), it is found that when a0>0 the normalized intensity correlation function C(s) increases with the increase of \tau0 or \tau2, and with value of \tau0 or \tau2 becoming larger, C(s) comes to saturation. With increasing the self-correlation time \tau1 of the additive noise, a minimum and a maximum will appear on curve of C(a) as a0>0. If a0<0, C(s) decreases with the increase of \tau0, \tau1, and \tau2.

A single-mode laser system with colored cross-correlated additive and multiplicative noise terms is considered. By the means of projection operator method, we study the effects of the cross-correlation time 'tau' and the cross-correlation intensity 'lambda' between noises on the normalized intensity correlation function C(s). It is found that if 'lambda'>0 ('lambda'<0), the normalized intensity correlation function C(s) increases (decreases) with increasing the cross-correlation time 'tau', and at large value of 'tau', the variation of the normalized intensity correlation function C(s) becomes small. With the increase of the net gain a0, C(s) exhibits a maximum when 'lambda' is larger. However, a minimum and a maximum appear on C(s) curves with the increase of a0 when 'lambda' becomes smaller and smaller.