ID 原文 译文
21645 相关滤波算法容易受到形变、运动模糊、相似背景等因素的干扰，导致跟踪任务失败。 Tracking effects of algorithms using correlation filter are easily interfered by deformation, motion blur and background clustering, which can result in tracking failure.
21646 为了克服以上问题，该文提出一种基于全局背景与特征降维的视觉跟踪算法。 To solve these problems, a visual tracking algorithm based on global context and feature dimensionality reduction is proposed.
21647 该算法首先提取紧邻目标的图像区域作为负样本供分类器学习，以抑制相似背景的干扰； Firstly, the image patches uniformly around the target are extracted as negative sample, and thus the similar background patches around the target are suppressed.
21648 然后提出一种基于主成分分析的更新策略，构建降维矩阵压缩HOG特征的维度，在更新分类器的同时减少其冗余度； Then, an update strategy based on principal component analysis is proposed, constructing the matrix to reduce the dimensionality of HOG feature, which can reduce the redundancy of feature when it updates.
21649 最后加入颜色特征表征运动目标，并根据特征对系统状态的响应强度进行自适应融合。 Finally, the color features are added to represent the motion target and the response of the system states are adaptively fused according to the features.
21650 在标准数据集上将该文提出的算法与Staple, KCF等其他算法进行了仿真对比，结果表明该文算法具有更强的鲁棒性，在形变因素的影响下，所提出的算法与Staple和KCF算法相比距离精度分别提升8.3%和13.1%。 Experiments are performed on recent online tracking benchmark. The results show that the proposed method performs favorably both in terms of accuracy and robustness compared to the state-of-the-art trackers such as Staple or KCF. When deformation occur, the proposed method is shown to outperform the Staple tracker and KCF algorithm by 8.3% and 13.1% respectively in median distance precision.
21651 深度学习在高维特征向量的信息提取和分类中具有很强的能力，但深度学习训练时间也比较长，超参数搜索空间大，从而导致超参数寻优较困难。 Deep learning has a strong ability in the high-dimensional feature vector information extraction and classification. But the training time of deep learning is so long that the optimal hyper-parameters combination can not be found in a short time.
21652 针对此问题，该文提出一种基于受限玻尔兹曼机(RBM)专家乘积系统的改进方法。 To solve these problems, a method of product of experts system based on Restricted Boltzmann Machine (RBM) is proposed.
21653 先将专家乘积系统原理与RBM算法相结合，采用全是真实概率值的参数更新方式会引起模型识别效果不理想和带来密度问题，为此将其更新方式进行改进； The product of experts theory is combined with the RBM algorithm and the parameter updating way is all adopted the probability value, which leads to the undesirable recognition effect and slightly worse density models, so the parameter updating way is improved.
21654 为加快网络收敛和提高模型识别能力，采取在RBM预训练阶段和微调阶段引入不同组合方式动量项的一种改进算法。 An improved algorithm with momentum terms in different combinations is used not only in the RBM pre-training phase but also in the fine-tuning stage for both classification accuracy enhancement and training time decreasing.