一种基于数据融合的机动目标跟踪预测算法

Forecast Algorithm for Tracking Moving Targets Based on Data Fusion

在机动目标的运动跟踪中,为了减小搜索区域,需要对下一时刻目标位置进行预测。本文以医疗康复领域中基于视频的运动分析为背景,针对二维平面中静止背景的运动目标提出一种基于数据融合的预测跟踪方法。首先采用多项式拟合算法和基于"当前"模型的改进卡尔曼滤波算法分别对运动目标进行位置预测,然后采用数据融合的方法得到最终的预测结果。最后用计算机仿真和实验对所提出的预测算法进行了验证,结果表明本文算法与多项式拟合和卡尔曼滤波算法相比,预测误差更小,跟踪精度更高。

In order to reduce the search region, it is necessary to predict the position of the moving object for the next time step in the neighbourhood of tracked target. On account of the video-based motion analysis used in the medical rehabilitation, a forecast algorithm based on data fusion for tracking moving targets in two-dimensional plane with the static background was proposed. Firstly, by means of the polynomial fitting and improved Kalman filtering algorithms based on the ＂current＂ mode, two predicted preliminary positions of a moving target were separately obtained. Then the resulting position was predicted by means of data fusion on two preliminary positions. Moreover, the proposed algorithm was tested by the computer simulation and the experiment. The results show that the proposed algorithm forecasts the position of moving targets more accurately than two other prediction methods based on polynomial fitting and improved Kalman filtering respectively.