We also establish a unified map frame to solve the super - resolution issue and pattern matching issue simultaneously . the frame has two advantages : first , the prior knowledge of the patterns is introduced into the super - resolution issue , and benefits the restoration of high - resolution image . second , more information of the low - resolution sequence is used in the pattern matching 在最大后验概率框架下建立了统一求解超分辨率与模板匹配问题的算法框架，它具有以下特点：在超分辨率中引入了模板库中的先验知识，提高了超分辨率的恢复效果；更充分利用了图像序列中的信息，提高了模板匹配的准确性。
Some new ideas are proposed in this thesis based on svm and ica : firstly , a modified svm method based on posteriori probability theory is given , which makes the classification super plane corrected from the original one . a better classification result is obtained without finding the best quadric optimization algorithm and large scale training datasets are reduced to small scale training datasets at the same time . secondly , ica is applied to the preprocessing period of the recognition character images for purpose of feature extraction and dimension reduction 本文在系统研究svm和ica的基础上提出了以下新的观点：其一是采用了引入后验概率的修正svm方法，它在原分类超平面的基础上不断修正分类超平面，提高分类正确率，从而避免了寻找最优二次规划的麻烦，同时将大规模训练样本集化为小规模训练样本集；其二是应用独立分量分析ica对需要进行识别的字符图像预处理，提取字符特征，降低输入数据的维数，从而可以为下一步的svm识别过程提供好的数据集，用以提高识别率和识别速度。
Monte carlo is a method that approximately solves mathematic or physical problems by statistical sampling theory . when comes to bayesian classification , it firstly gets the conditional probability distribution of the unlabelled classes based on the known prior probability . then , it uses some kind of sampler to get the stochastic data that satisfy the distribution as noted just before one by one 蒙特卡罗是一种采用统计抽样理论近似求解数学或物理问题的方法，它在用于解决贝叶斯分类时，首先根据已知的先验概率获得各个类标号未知类的条件概率分布，然后利用某种抽样器，分别得到满足这些条件分布的随机数据，最后统计这些随机数据，就可以得到各个类标号未知类的后验概率分布。
It is a markov network , we confirm the transfer rule of this network based on two kind of relationship , one is the image and the scene , the other is the current scene and its neighbour . learning parameters of this network come from training examples , we can obtain a exact local maximum of the posterior probability for the scene , thereby we generate a effective edge detection result for original blurry importing image 它的体系结构是一个马尔可夫网络，根据图像与景物、景物与景物之间的联系来确定网络上的信息传递规则，并从大量的事例中学习这些网络参数，可以高效地为所求景物找到一个精确的后验概率的局部最大值，从而为原模糊图像获得高效的边缘标记结果。
Among others , the probability analysis approach has difficulty in deciding objective probability , and thus it is necessary to obtain subjective probability through expert empirical prediction , modify it by the bayesian formula and get a posteriori probability , and substitute it for objective probability in risk measurement and risk premium calculation 其中，概率分析方法在应用中就存在客观概率不易确定的难点问题，因此需用专家经验预测法得到主观概率后，利用贝叶斯公式加以修正并获得后验概率，再用后验概率代替客观概率进行风险的度量及风险收益的计算。
On this basis a nonlinear filtering technique of sequential monte carlo particle filter based on bayesian approach is emphatically disussed which the posterior distribution of the state variables can be represented by a set of weighted particles , so the method base advantages over the above algorithms in robustness and accuracy for nonlinear non - gaussian filtering problems 在此基础上重点论述了一种基于贝叶斯原理的序贯蒙特卡罗粒子滤波技术，该方法通过粒子的加权和表征后验概率密度，获得状态估值，在处理非线性非高斯系统的状态估计问题时精度逼近最优，鲁棒性更好。
Applications of multiple - model smoothing algorithms for maneuvering target tracking are studied via simulation , some important conclusions are obtained . based on model - set sequential likelihood ratio , an enhanced agimm , in which model - set adaptation is implemented by jointly utilizing model posterior probability and predication probability , is proposed , simulation results indicate that improvements of both dynamic and steady state tracking performance are achieved with the enhanced algorithm 仿真研究了多模型平滑算法在机动目标跟踪中的应用；利用模型集合序贯似然比检验，提出了一种综合利用模型后验概率和预测概率实现模型集合自适应的综合格自适应多模型算法，仿真实验表明算法有效改善了动态跟踪精度和稳态跟踪性能。
There are mainly two type of algorithms used for spatial spectrum estimation : one is those based on bayesian maximum likelihood method , like the ml ( maximum likelihood ) algorithm , maximum entropy method and etc . , the others are based on the spatial decomposition or projection of correlation matrix , this kind of algorithm include vector characterization method , music ( multiple signal classification ) algorithm , projection matrix method , etc . music is a classical spatial spectrum estimation algorithm that has a super high resolution and is widely used today , however , it cannot estimate doa of signals that are correlated 空间谱估计的算法大致分两大类:一是基于极大似然估计和最大后验概率估计统计理论的算法,包括:极大似然估计法( ml ) 、最大熵法等;另一类是基于对协方差矩阵进行子空间分解或投影的算法,包括:矢量特征法、多重信号分类法( music ) 、投影矩阵法等。其中, music法是一种经典的空间谱估计主流算法,具有超强的分辨性能,但它无法实现对相干信号进行测向分辨。
Edges of the image are detected out firstly , labeled according to the motion that they obey then and the areas of the frame between edges are divided into regions . at last , using the bayesian framework presented determines the most likely region labeling and depth ordering with the labeled edges 首先使用经典的canny算子检测出一帧图像的边缘，然后对其进行运动估计、边缘和区域标定，再应用最大后验概率的贝叶斯方法搜索出不同区域的极大似然分割，给出不同运动层的相对深度标定。
The local structure of chain - type models and tree models based - on bayesian dirichlet metric are mainly studied , and the theoretic foundation for creating those models is provided . simultaneously , the algorithms to create chain - type and tree models are presented . these chain - type and tree models to be investigated have obvious causality with high posterior probabilities 本文通过研究基于bd度量的链形和树形模型结构性质，给出了构建链形和树形模型的算法步骤，使所构建的图形模型具有明显的因果关系和较大的后验概率。
