摘要
提出一种新的FAST-Snake目标跟踪方法,利用改进的FAST角点特征匹配来估计目标轮廓在帧间的全局仿射变换,将投影轮廓点作为Snake模型的初始化轮廓.为提高跟踪实时性,在Snake能量模型中定义了先验约束能,并用限定搜索方向的贪婪算法(Greedy algorithm)实现局部轮廓优化.实验包括三维目标数据库及真实场景视频,验证了提出方法的均方误差(Mean square error,MSE)及收敛速度评估均优于对比算法,并具备对复杂运动及局部遮挡的适应能力.
We present a novel FAST-Snake tracking approach using improved FAST-feature matching to estimate affine transform of contour points between frames as the initial contour of the Snake model. For real-time tracking, we define a prior constraint energy in the Snake model and adopt the greedy algorithm to implement contour optimization. Experi- ments involving 3-D object database and video sequences show that the proposed approach is superior to its counterpart in terms of mean square error （MSE） and convergence speed, and that it has the adaptability to complex motion and partial occlusion.
出处
《自动化学报》
EI
CSCD
北大核心
2014年第6期1108-1115,共8页
Acta Automatica Sinica
