【摘要】 提出一种结合多种图像特征,在多摄像机环境下跟踪人体运动的方法.通过定义人体模型、摄像机投影模型以及相似性度量模型来得到优化框架下的目标函数,并使用牛顿-高斯优化算法对其进行求解.模拟数据和实际数据的实验表明,文中方法比仅仅使用灰度特征,跟踪结果得到了改善,对比实验结果也优于基于概率算法的退火粒子滤波.更多还原

【Abstract】 We present a video-based 3D body tracking method from multiple cues in multiple cameras environment.By defining the body model,image projection model and similarity measurement model,the tracking object function is proposed,then resolved using the Newton-Gaussian method.The experimental results on simulated data and the real data show the tracking is improved by combining several image cues,and the tracking performance is better than the annealed particle filter algorithm.更多还原