【摘 要】
：
提出一种多任务稳健主成分分析方法,用以结合多视觉特征实现运动目标分割.给定由多类型特征矩阵描述的视频数据,将它分解为低秩和稀疏部分,其中的稀疏部分对应于运动目标.该矩阵分解过程是一个凸优化问题,通过用ALM方法最小化核范数和2,1-范数的约束组合.与仅利用单类型特征的方法相比,本文提出的方法能够结合多类型特征,因此可获得更加精确可靠的结果.对HumanEva和Change Detection两个数
论文部分内容阅读
提出一种多任务稳健主成分分析方法,用以结合多视觉特征实现运动目标分割.给定由多类型特征矩阵描述的视频数据,将它分解为低秩和稀疏部分,其中的稀疏部分对应于运动目标.该矩阵分解过程是一个凸优化问题,通过用ALM方法最小化核范数和2,1-范数的约束组合.与仅利用单类型特征的方法相比,本文提出的方法能够结合多类型特征,因此可获得更加精确可靠的结果.对HumanEva和Change Detection两个数据集的实验表明了该方法的有效性.
其他文献
We propose a novel method by combining the total variation(TV) with the high-degree TV(HDTV) to improve the reconstruction quality of sparse-view sampling photoacoustic imaging(PAI). A weighing functi
为了客观、准确及定量地评估激光干扰效果,给出了图像边缘强度的定义及计算方法,提出了一种基于边缘强度相似度的干扰图像尺度.该尺度通过计算目标图像和背景图像的相似程度来衡量激光干扰效果,相似度越高表明激光干扰效果越好.激光视场内及视场外干扰实验与数值计算表明:该图像尺度可以较好反映激光干扰对CCD成像探测系统目标检测算法性能的影响.同时,通过对荷兰TNO研究所提出的Search_2数据库的计算表明,提
2014年8月3日16时30分,云南鲁甸突发6.5级地震。这一刻,山崩地裂,房屋垮塌,生命消殒。人民子弟兵闻令而动。不惜一切代价向震中突进,抢救生命、抗击地震灾害战斗随即打响。8月4日早上.正在上班途中的我们接到去一线采访命令,迅即改道奔赴首都机场,直飞昆明,而后驱车奔赴鲁甸。当晚11点,顺利到达鲁甸
After the Wenchuan MS 8.0 Earthquake, which occurred on May 12, 2008, in Sichuan Province, China, we conducted a series of hydraulic fracturing stress measurements in three 200 m deep boreholes(ZK01,
The Lushan Earthquake induced a large number of geohazards. They are widely distributed and caused serious damages. The basic characteristics, formation mechanisms and typical cases of geohazards indu
On the basis of elastic rebound theory,using the horizontal velocity field of the Chinese mainland calculated from GPS data during three observation periods from 1999 to 2007,the velocity components t
The deep structure features around Tancheng M8(1/2)earthquake occurred in 1668 have been analyzed. Based on the crustal velocity structure obtained from travel-time tomographic inversion,especially th
Tidal phase of water temperature in Qixian well,Shanxi Province is ahead of water level,which is different from the normal tidal relationship between water temperature and water level. Observation cur
A trend increase in apparent resistivity has been observed in the N30°E monitoring direction at Garzê Seismic Station since July 2011. This increase trend in geo-electric resistivity has been observed
针对传统的基于颜色特征的cam shift车辆跟踪算法易受到光照变化的干扰而无法有效地跟踪动态目标的问题,提出一种多特征融合的cam shift车辆跟踪算法.将局部二值模式(local binary pattern,LBP)纹理特征改进为局部三值模式(local ternary pattern,LTP)纹理特征,与HSV(hue,saturation,value)颜色特征相结合,建立目标模型,对运