ID 原文 译文
26305 结果表明综合识别方法和仅基于窄带或宽带极化特征的识别方法相比具有更好的目标识别性能。 Results showed that the comprehensive identification method and only based on narrowband or broadband polarization feature recognition method has better performance of target recognition.
26306 针对尺度不变特征变换(scale invariant feature transform，SIFT)算法在特征点匹配时容易出现误匹配现象，提出了一种基于区域重叠核加权Hu矩的SIFT误匹配点剔除算法。 In view of the scale invariant feature transform (scale invariant feature transform, SIFT) algorithm is prone to error when matching feature points matching, is proposed based on a regional overlap nuclear weighted Hu moment SIFT out matching point algorithm.
26307 该算法首先通过对SIFT描述子区域内的重叠4邻域计算Hu矩，生成能够描述纹理特征与轮廓特征的种子点描述子； The algorithm firstly through overlapping of SIFT descriptor area 4 neighborhood Hu moment calculation, generate seeds can describe the texture characteristics and contour feature point descriptor;
26308 其次，根据描述子的区域特点利用核函数对种子点描述子进行加权，生成63维区域重叠核加权Hu矩描述子； Secondly, based on the description of regional characteristics of the seed point descriptor to make use of kernel function is weighted, generate 63 d zone overlapping nuclear weighted Hu moment descriptor;
26309 最后用巴氏(Bhattacharyya)系数计算归一化后描述子的相似度，并剔除相似度较小的匹配点。 With pap finally, Bhattacharyya coefficient calculation after normalization descriptor similarity, smaller and eliminate similarity matching points.
26310 将该算法与其他3种算法进行对比，实验结果表明，该算法的鲁棒性最强，实时性较高，综合性能最优。 Compare the algorithm with the other three kinds of algorithms, the experiment results show that the robustness of the algorithm is the strongest, high real-time performance and comprehensive performance of the optimal.
26311 提出了一种基于非下采样Contourlet变换(nonsubsampled Contourlet transform，NSCT)的红外与可见光图像融合方法。 Put forward a kind of based on the next sampling Contourlet transform (nonsubsampled Contourlet transform, NSCT) of infrared and visible light image fusion method.
26312 首先对原红外图像进行图像分割，确定目标区域与背景区域，并将其映射到可见光图像中；然后对红外和可见光图像进行多尺度、多方向分解，分解后的低频部分在目标区域选择红外图像低频系数、在背景区域选择可见光图像低频系数，高频部分使用方向方差加权信息熵最大作为融合策略进行融合；最后对融合的系数进行重构得到融合图像。 First of all, the original infrared images for image segmentation, target region and background region, and map it to visible light image;The infrared and visible light image and multi-scale decomposition, multiple directions, the low frequency part of the decomposed in the target area for low frequency coefficient of infrared image, in the background region selection visible light image low frequency coefficient, high frequency part use direction, the maximum variance weighted information entropy fusion as a fusion strategy;Finally, the fusion coefficient of refactoring get fused images.
26313 实验结果表明，本文算法在保留图像细节信息、增加信息量、方便目标检测方面都有显著地提高。 The experimental results show that the algorithm in keep the details of image, increase the amount of information, convenient detection has significantly improved.
26314 关键状态是系统风险演化中的重要环节，识别关键状态对系统风险控制等具有重要意义。首先结合动态事件树基本方法给出了一类关键状态和关键事件的数学定义，然后提出了此类关键状态的两种搜索算法——基于子树分解的搜索(sub-tree decomposition，STD)算法和基于逻辑运算的搜索(boolean calculation，BC)算法，并对其计算量进行了对比分析。 Critical state is an important link of system evolution of risk, identify the critical state is of great importance in the system of risk control, etc. First with basic dynamic event tree method gives a class key status and key mathematical definition of events, and then puts forward such key state of the two kinds of search algorithms, based on the decomposition of the subtree search (sub - tree decomposition, STD) algorithm and search based on logical operations (Boolean calculation, BC) algorithm, and has carried on the comparison and analysis on its amount of calculation.