基于有效特征点的运动目标匹配跟踪算法

Moving target matching and tracking algorithm based on effective feature points

在实际场景下,由于目标快速运动,遮挡、复杂背景等因素影响,使得对运动目标的跟踪效果较差。为了解决上述问题,提出了一种基于有效特征点的自适应目标跟踪算法,实现对运动目标的稳定实时跟踪。首先对光流法追踪及特征点匹配得到的稳定特征点进行融合,得出能有效代表目标的特征点。然后利用上述特征点求出目标的角度变化和尺度变化,进而实现对目标的跟踪。其中,在特征点匹配过程中,采用马尔科夫方向预测器求出目标框在下一时刻中的大概位置,缩小了检测区域并增强了对相似目标的辨识能力。实验表明,该方法具有较高的跟踪精度、处理速度和鲁棒性,尤其是当目标快速运动或者发生形变时,能够有效的减少中心错误率。

In the actual situation, due to the target fast moving , occlusion , complex background and other factors ,making the tracking results for moving target are poor. In order to solve the above problems, we proposed an adaptive target tracking algorithm based on effective feature points to achieve stable real- time tracking of ground rigid targets. Firstly fusing the stable feature points which were tracked by optical flow and obtained after feature points matching .After fusing we obtained the feature points which can effectively represent the target .Using the above feature points to calculate the target＇ s angle and scale changes, and then achieves the target tracking .In the process of feature points matching, using Markov direction predictor to find the approximate position of the target frame at the next moment which can narrow the detection area and improve the recognition ability of the similar target. The experiments show that the method has high tracking precision, processing speed and robustness. And it can effectively reduce the tracking error rate, especially when the target moves fast or occurs deformation.