【摘要】 针对Kalman预测跟踪和K-近邻数据关联算法的优缺点,研究一种基于Kalman预测和K-近邻的多目标跟踪方法。该方法首先利用Kalman滤波预测出运动目标在下一帧中最可能出现的位置,接着根据当前帧目标位置和预测目标位置的距离,确定搜索半径,利用K-近邻数据关联算法,在该半径范围内,计算与预测点欧式距离最短的目标,并将其确定为真实目标位置。在MATLAB仿真环境下实现该跟踪算法,实验结果表明,用该方法进行多目标跟踪时,跟踪效果和性能较为稳定和可靠。此外选择合理的K值,能减少运算量,加快系统处理速度。更多还原

【Abstract】 In this paper,directing at the strong points and weak points of Kalman filter based on tracking method,a novel approach to tracking of muti-objects is studied.By using Kalman-filter,the researchers predict locations where objects most probably appear in a next-frame,determine the search radius and k,calculate the shortest Euclidean distance with the predicted target object in this radius by using k-nearest neighbor,and then determine the true target location.Based on the MATLAB simulation environment to achieve the tracking algorithm,experimental results show that tracking results and performance is better.In addition,selecting a reasonable k values,can reduce the computation and speed up the system processing speed.更多还原