基于支持向量描述的自适应高光谱异常检测算法

提出了一种应用于高光谱异常检测的自适应支持向量数据描述方法.根据高光谱数据和局部异常检测模型的特点,通过局部背景分波段二阶分布统计,分析了核参量与局部背景总体标准差的变化关系,构造了随检测背景变化的局部检测核参量,使得检测算法针对不同背景分布自适应地调整检测核参量.克服了传统支持向量描述算法由于采用固定核参量带来的复杂背景下检测性能下降的问题.通过模拟数据和真实高光谱数据的测试检验,接收机特性曲线表明该算法相对于传统固定核参量支持向量数据描述方法,在相同虚警概率下检测概率提高了10%.     

二维阵列光场衍射自成像效应的倒格矢分析

基于衍射的角谱理论,分析了一般二维周期阵列光场的衍射特性.提出了一种用阵列光场的倒格矢来研究和描述衍射自成像（或泰伯效应）的方法.给出了用倒格矢表示的一般衍射自成像条件.在此基础上,对典型的非正交六方型二维周期阵列光场的泰伯效应和分数泰伯效应进行了定量分析和计算机模拟.     

基于支持向量机的模型传递方法研究

模型传递是以数学方法通过在2台不同仪器之间寻求一种变换关系来增强光谱仪数据通用性、可比性的一种基本途径。由于实际测量数据具有非线性特征,加上校正样本集合的有限性,使得解决小样本条件下非线性关系的模型传递问题显得尤为重要。文章在概述支持向量机基本原理的基础上,探讨了支持向量机方法在光谱仪的模型传递问题中的应用,提出了基于支持向量机的分段直接校正方法,最后采用计算机模拟的方式对该方法进行了举例说明,并和人工神经网络方法进行了相应的比较。     

混合气体红外光谱支持向量机分析的新方法

介绍了一种基于支持向量机的混合气体红外光谱组分浓度和种类分析的新方法。利用核函数将组分气体特征吸收谱线重叠严重的混合气体光谱在高维空间变换后,建立SVM回归校正模型,进行混合气体浓度分析。在利用支持向量机回归校正模型进行混合气体组分浓度分析的同时,证明支持向量机回归校正模型也可用于混合气体组分种类分析。对不同组分和不同组分浓度的混合气体红外光谱数据进行了实验,研究了谱仪扫描间隔、分析特征波长范围、核函数和惩罚因子等因素对分析结果的影响。混合气体组分浓度实验结果的最大平均绝对误差Mean AE为0.132%;混合气体组分种类识别的准确率大于94%。解决了传统的光谱分析方法中光谱特征谱线重叠、光谱数据的维数大、定性和定量分析无法使用同一方法等问题,可用于其他混合气体的红外光谱分析,具有实际应用价值。     

激光陀螺中背向散射的研究

为了研究谐振腔形状变化对激光陀螺中背向散射的影响,基于矢量叠加理论,分析了激光陀螺背向散射源及其对总的背向散射的影响.根据背向散射对顺逆时针光的耦合作用,推导出了总的背向散射系数与顺逆时针旋转光光强的关系式.基于这一关系,设计实验,通过施加电压使谐振腔反射镜到不同位置来改变光腔形状变化,同时测量光强交流量的振幅,得到了总的背向散射系数的变化曲线.通过与仿真结果对比,验证了背向散射的矢量叠加理论.     

Optimising time-varying gradient orientation for microstructure sensitivity in diffusion-weighted MR

Here we investigate whether varying the diffusion-gradient orientation during a general waveform single pulsed-field gradient sequence improves sensitivity to the size of coherently oriented pores over having a fixed orientation. The experiment optimises the shape and the orientation of the gradient waveform in each of a set of measurements to minimise the expected variance of estimates of the parameters of a simple model. A key application motivating the work is measuring the size of axons in white matter. Thus, we use a two compartment white matter model with impermeable, single-radius cylinders, and search for waveforms that maximise the sensitivity to axon radius, intra-cellular volume fraction and diffusion constants. Output of the optimisation suggests the only benefit of allowing the gradient orientation to vary in the plane perpendicular to the cylinders is that we can gain perpendicular gradient strength by maximising two orthogonal gradients simultaneously. This suggests that varying orientation in itself does not increase the sensitivity to model parameters. On the other hand, the variation in a plane containing the parallel direction increases the sensitivity significantly because parallel sensitivity improves the diffusion constant estimates. However, we also find that similar improvement in the estimates can be achieved without optimising the orientation, but by having one measurement in the parallel and the rest in the perpendicular direction. The optimisation searches a very large space where it cannot hope to find the global minimum so we cannot make a categorical conclusion. However, given the consistency of the results in multiple reruns and variations of the experiments reported here, we can suggest that for probing coherently oriented systems, pulse sequences with variable orientation, such as double-wave vector sequences, do not offer more advantage than fixed orientation sequences with optimised shape. The advantage of varying orientation is however likely to emerge for more complex systems with dispersed pore orientation.     

Equilibrium polymerization in a solvent: Solution on the Bethe lattice

The lattice model for equilibrium polymerization in a solvent proposed by Wheeler and Pfeuty is solved exactly on a Bethe lattice (core of a Caylay tree) with general coordination numberq. Earlier mean-field results are reobtained in the limitq, but the phase diagrams show deviations from them for finiteq. Whenq=2, our results turn into the solution of the one-dimensional problem. Although the model is solved directly, without the use of the correspondence between the equilibrium polymerization model and the diluten0 model, we verified that the latter model may also be solved on the Bethe lattice, its solution being identical to the direct solution in all parameter space. As observed in earlier studies of the puren0 vector model, the free energy is not always convex. We obtain the region of negative susceptibility for our solution and compare this result with mean field and renormalization group (-expansion) calculations.     

波矢空间中电磁场的几何相位

将电磁场作Fourier变换可把Maxwell方程组转写成为波矢空间中的Schrodinger型方程。然后采用Aharonov和Anandan关于态矢循环演变时相位变化的理论给出了波矢空间中电磁场的相位.其中包括几何相位。     

Differentiable conjugacy of the Poincaré type vector fields on

We prove that on , except for those germs of vector fields whose linear parts are conjugated to , any two Poincaré type vector fields are at least conjugated to each other provided their linear approximations have the same eigenvalues and the nonlinear parts are generic.