ID	原文	译文
2483	实验表明，该网络模型可以完成多类型目标的显著性检测，具有通用性好，准确度高等优点。	The experiments show that the proposed algorithm can accomplish the saliency detection of various targets, which has the ad-vantages of good universality and high accuracy.
2484	如何有效地抵抗组合几何攻击，并且提高水印的嵌入容量，一直是数字水印领域一个具有挑战性的课题。	How to resist combined geometric attacks effectively and improve the embedded capacity of the watermark has always been a challenging topic in the field of digital watermarking.
2485	针对此种状况，本文提出了一种抗组合几何攻击，且嵌入容量较大的水印算法。	In view of this situation, a watermarking algorithm that resists combined geometric attacks and has a larger embedded capacity is proposed in this paper.
2486	首先利用 SURF 算法从受攻击的图像提取特征点，并与原图像中的少量鲁棒性较强的特征点匹配，根据特征点对间的坐标关系准确估计出仿射矩阵;	Firstly, feature pointsfrom the attacked image are extracted with SURF algorithm, then they are matched with a small amount of feature points ofstrong robustness, which are extracted from the original image. The matching point pairs are used to estimate the affine ma-trix accurately and then the geometric attacks are corrected to resynchronize the embedded watermarks.
2487	然后根据仿射矩阵对失真的图像进行几何校正，恢复水印的重同步。	Due to positioning er-rors introduced during the attacking /correction process, some positioning watermarks are embedded in the spatial domain toimprove the accuracy of the resynchronization.
2488	由于几何攻击与还原过程中的像素插值仍会导致变换域水印的位置发生细微的误差，本文在空间域嵌入一层定位水印以便精准地同步;最后，考虑到 LT 码具有良好的纠删性能，因而先将变换域水印信息进行 LT 码编码，以尽量提高算法的抗剪切性能。	Considering the excellent erasure correcting performance of LT codes, the wa-termark bit sequence is encoded by LT-code such that the anti-cropping performance of the algorithm can be improved.
2489	实验结果表明，该算法不仅实现了较高容量的嵌入，而且对于剪切、组合几何攻击具有较好的鲁棒性。	Theexperimental results show that the proposed algorithm not only has a larger embedding capacity, but also is resistant to the cropping and combined attacks.
2490	对社会化媒体产生的大量短文本进行聚类分析具有重要的应用价值，但短文本往往具有噪音数据多、增长迅速且数据量大的特点，导致现有相关算法难于有效处理。	Clustering a large number of short texts in social media has great value in applications. However, short texts often have these characteristics:lots of noises, growing rapidly and massive data. Most existing short text clustering al-gorithms are not effectively enough to process such short texts.
2491	提出一种基于增量式鲁棒非负矩阵分解的短文本在线聚类算法 STOCIRNMF。	Aiming at this problem, we propose an algorithm of short text online clustering based on incremental robust nonnegative matrix factorization (STOCIRNMF).
2492	STOCIRNMF 基于非负矩阵分解构建短文本聚类模型，通过 l2，1范数设计模型的优化求解目标函数提高鲁棒性，同时应用增量式迭代更新规则实现短文本的在线聚类。	This algorithm uses NMF tobuild the short text clustering model and applies l2, 1norm to devise its objective function for improving its robustness. Mean-while, STOCIRNMF can cluster short texts incrementally by using incremental iterative update rules.