ID 原文 译文
25275 在众包学习中，使用标记集成算法得到的集成标记中仍然存在一定程度的标记噪声。 In crowdsourcing learning, a certain level of label noise still exists in integrated labels obtained by employing ground truth inference algorithms．
25276 本文受三重训练思想的启发，提出了一种基于 tri-training 的众包标记噪声纠正算法(Tri-Training-based Label Noise Correction，TTLNC)。 Inspired by the tri-training idea, this paper proposes a tri-training-based label noise correction (TTLNC) algorithm for crowdsourcing.
25277 TTLNC 首先使用过滤器获得干净集和噪声集，然后在干净集上进行 bagging 分别训练三个不同的分类器，并通过这些分类器重新标注噪声集中的实例，同时按照实例分配策略将实例分配给相应的训练集。 TTLNC at first employs a filter to get a clean set and a noisy set and then trainsthree different classifiers from the bagged clean set. Furthermore, each instance from the noisy set is relabeled by these classifiers and assigned to the corresponding training set according to the designed instance assignment strategy．
25278 最后在新训练集上重新训练三个不同的分类器，并用新分类器的分类结果重新标注所有实例。 Finally, three classifiers are retrained on three new training sets and are used to relabel all instances．
25279 在仿真标准数据和真实众包数据集上的实验结果表明TTLNC 比其他四种最先进的噪声纠正算法在噪声比和模型质量两个度量指标上表现更优。 Experimental results on both simulated benchmark data and realworld crowdsourced data show that TTLNC significantly outperforms other four state-of-the-art noise correction algorithms in team of the noise ratio and the model quality．
25280 针对小型移动机器人对人体目标快速运动或遮挡导致的跟踪准确率降低甚至跟踪失败问题，通过建立足部运动模型预测双脚位置信息，获得核相关滤波(KCF，Kernel Correlation Filter)目标检测区域，再结合输出响应峰值邻域相关检测，提出了运动模型引导的自适应核相关滤波算法。 Aiming at the problem of low tracking accuracy and even tracking failure caused by fast motion or occlusion of human targets by small mobile robots, a foot motion model was established to predict the position informationof feet, and the target detection region of kernel correlation filter (KCF) was obtained. A motion model guided adaptive kernel correlation filtering algorithm is proposed by combining with the output response peak neighborhood correlation detection．
25281 对实际拍摄的七组不同情况下的视频进行了足部目标跟踪实验，结果表明运动模型引导的自适应响应 KCF 算法平均跟踪准确率最高，且在短时间遮挡情况下的算法跟踪准确率也达到 86% ，明显高于自适应响应 KCF、BACF(Background Aware Correlation Filters)以及 SAMF(Scale Adaptive kernel cor-relation filters with Multiple Features)三种跟踪算法。 Foot tracking experiments were carried out on seven groups of videos under different scenarios. The results show that the average tracking accuracy of the adaptive response KCF algorithm guided by the motion model is the highest, and the tracking precision rate of the algorithm reaches 86% in the case of shortterm occlusion, which is significantly higher than that of the adaptive response KCF, BACF and SAMF algorithm．
25282 最后在 ＲOS(Ｒobot Operating System)下将所提算法应用于 Turtlebot机器人目标跟踪测试，成功克服了遮挡情况对足部跟踪带来的影响，验证了所提算法具有较强的鲁棒性和实时性。 Finally, the proposed algorithmis applied to the target tracking test of a Turtlebot robot under the ＲOS (Ｒobot Operating System), which successfully overcomes the influence of occlusion on feet tracking, and verifies that the proposed algorithm has strong robustness and real-time performance．
25283 针对传统降水粒子分类算法存在的过度依赖专家经验和模型预设误差问题，本文提出了一种基于离散属性贝叶斯网络(Bayesian NeTwork，BNT)的双偏振气象雷达降水粒子分类(Hydrometeor Classification，HC)方法。 The over-reliance on expert experience and model preset errors in traditional precipitation particle classification algorithms are discussed. This paper proposes a dual-polarization hydrometeor classification (HC) method based on discrete attribute Bayesian NeTwork (BNT)．
25284 首先对双偏振气象雷达获取的偏振参量取值进行离散化处理生成离散化标准，并根据离散化标准制作训练数据集合; Firstly, the value of polarization parameters obtained by the dual-polarization meteorological radar is discretized to generate a discretization standard, and the training data set is made according to the discretization standard．