摘要
运动目标跟踪的根本任务是根据目标的运动模型和图像特征估计它们的轨迹。提出一种运动目标检测、跟踪的方法。首先使用基于自适应混合高斯模型的背景差方法提取运动区域。目标的运动估计采用扩展卡尔曼滤波,由预测位置确定初始的候选区域。然后根据目标与候选区域的变化程度确定匹配需要的特征信息。如果目标只有一个候选区域并且它们之间的区域特征变化微小,那么它们的匹配不需要额外的信息。如果目标有多个候选区域或者单个候选区域可是它们的区域特征变化激烈,除了区域特征外还使用边缘特征,通过计算目标和候选区域的边缘的部分Hausdorff距离来确定目标的最佳匹配区域。实验结果表明,该方法在存在遮挡的情况下也能够连续的跟踪多个运动目标。
The basic task of tracking can be defined as the problem of estimating the trajectory of a target in the image plane as it moves in the scene. In this paper, a method is proposed for detecting and tracking moving targets. First, adaptive mixture of Gaussians is used to extract moving regions. Second, targets are tracked as a bounding rectangle, and extended Kalman filter is used to predict the position of bounding rectangle in the next frame. In the same time, which feature being used to match is determined through the region feature between target and the candidate. If the region feature changes slightly, the target can be tracked using the region feature. If the region feature changes acutely or the target corresponds to more than one candidate regions, then the target may be occluded. In this case, the target and the candidate regions are compared using partial Hausdorff distance. Experimental results show that the algorithm presented in this paper can continuously track targets.
出处
《指挥控制与仿真》
2008年第2期17-20,共4页
Command Control & Simulation
