摘要
本文提出了一种在相控阵雷达回波数据序列中用高斯混合体模型 (GMM)检测与跟踪运动目标的在线算法 .首先 .回波数据中的每一固定位置的历史数据用点过程描述 .然后用GMM逼近此点过程 .GMM参数随新数据的到来不断更新 .接着建立具有自适应特性的背景模型 .将每帧回波数据分割为背景和前景 .对已标记为前景的数据用连通分支进行分类 .求出目标的中心位置、大小、径向运动速度、角速度 .最后用卡尔曼滤波器对运动目标进行跟踪 .试验结果表明 .本文算法对于复杂场景中运动目标的检测与跟踪具有较好的鲁棒性和实时性 .具有较强的实用价值 .
An on-line real-time tracking algorithm for radar image sequences is proposed.First,a GMM (Gaussian mixture model) is used to approximate values of a particular pixel of the radar image sequences,and parameters of the GMM are updated each time.Then,we use a simple heuristic to hypothesize which Gaussian of the mixture is most likely to be part of the adaptive background model.After the background model is established,pixel values that don't match the pixel's 'background' Gaussians are grouped using connected components.Features of the targets such as center position,size,radial and angular velocities are also computed in the mean time.Finally,the connected components are tracked across frames using a Kalman filter-based tracker.The experimental result shows that the algorithm is robust in clutter and easy to implement on-line.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2003年第3期433-436,共4页
Acta Electronica Sinica
基金
国家创新研究群体科学基金资助项目 (No .60 0 2 4 30 1 )
