ID 原文 译文
20625 在特征匹配阶段，采用仿射模型模拟变换关系建立几何约束条件，克服SIFT算法由于忽略几何信息而产生的误匹配。 In feature matching phase, the affine transformation model isused to simulate the transformation relation and establish the geometric constraint, to overcome themismatching because of ignoring the geometric information.
20626 实验表明：该方法在匹配精度和实时性方面均优于SIFT算法，且对光照、模糊、尺度等变换具有良好的鲁棒性，能够更好地实现景象匹配。 The experimental results show that the proposedmethod is superior to the SIFT in efficiency and precision, also has good robustness to light, blur and scaletransformation, achieves scene matching better.
20627 针对机器人在家庭环境下的目标检测问题，该文提出一种基于动作注意策略的树形双深度Q网络(TDDQN)目标候选区域提取的方法，该方法将双深度Q网络(DDQN)的方法与树结构的方法相结合，通过执行改变检测框的动作以使目标逐渐集中在检测框内。 Considering the problem of object detection of robots in the home environments, a Tree-Double DeepQ Network (TDDQN) based on the attention action strategy is proposed to determine the locations of region proposals. It combines DDQN with hierarchical tree structure.
20628 首先采用DDQN方法在执行较少的动作后选择出当前状态的最佳动作，获取符合条件的候选区域。 First, DDQN is used to select the best action ofcurrent state and obtain the right region proposal with a few actions executed.
20629 然后根据执行所选择动作之后所得到的状态重复执行上述过程，以此构成树结构的多条“最佳”路径。 According to the state obtained after executing the selected action, the above process is repeated to create multiple "best" paths of the hierarchical tree structure.
20630 最后采用非极大值抑制的方法从多个符合条件的候选区域选择出最佳候选区域。 The best region proposal is selected using non-maximum suppression on region proposals that meet the conditions.
20631 在PascalVOC2007以及Pascal VOC2012上的实验结果表明，在不同数量的候选区域、不同阈值的IoU和不同大小以及不同种类对象的实验条件下，所提方法较其他方法都有着更好的检测性能，可以较好地实现目标检测。 Experimental results on Pascal VOC2007 and Pascal VOC2012 show that the proposed method based on TDDQN has better detection performance than other methods for region proposals of different numbers, different Intersection-over-Union (IoU) values and objects of different sizes and kinds, respectively.
20632 针对现有相干分布源直接定位方法中存在的依赖分布模型、计算复杂等问题，该文提出一种基于非圆信号特征的对称旋转不变直接定位算法。 The existing Direct Position Determination (DPD) algorithm of Coherently Distributed (CD) sources rely on the distribution model of CD sources with huge computation cost, which is not practical. To improve further the localization performance, a novel DPD algorithm of CD sources that profits from the characteristics of noncircular signals is proposed based on the symmetric shift invariance of the centrosymmetric array.
20633 该方法首先根据分布源参数化假设建立基于数据域信息的直接位置估计模型，并利用非圆信号特征扩展接收信号的协方差矩阵。 With the parameterization assumption of CD sources, the direct position determination model is firstly constructed by combining the characteristics of noncircular signals.
20634 然后针对中心对称阵列，证明了相干分布源的确定性角信号分布函数矢量具有对称特性，基于这一特征建立了扩展方向矢量的旋转不变关系；构造了融合多个观测站信息的目标函数，直接估计目标位置，避免了对分布模型的依赖，且降低了待估计参数维度。 Then, it is proved that for any centrosymmetric array,the generalized steering vector of CD sources has the property of symmetric shift invariance. Base on this characteristic, the positions of CD sources are directly estimated by fusing the information of all observation stations with no need to consider the distribution model, which reduces the dimension of the parameter to be estimated.