基于多新息Kalman滤波的TLD改进算法

Improved TLD Algorithm Based on Multi-innovation Kalman Filter

针对跟踪检测学习(tracking learning detection,TLD)跟踪算法中目标被遮挡后跟踪失败以及跟踪精度不高的问题,本文提出基于多新息Klaman滤波的TLD改进算法,在原始TLD跟踪算法的基础上加入了多新息Klaman滤波算法。改进算法对跟踪目标建模,将TLD跟踪算法的结果作为系统当前状态的观测值,结合多新息Kalman滤波算法的预测值,最优化检测结果,作为当前帧中目标的跟踪位置。通过实验对原始TLD和改进后的TLD算法进行比较,通过在标准测试序列的实验验证,加入多新息Kalman滤波的TLD改进算法与原始TLD算法相比,其跟踪误差更小,而且实现了对跟踪目标被遮挡后的位置预测。

Abstract： To solve the problems of tracking failure caused by blocked target and low tracking precision in the tracking learning detection （TLD） algorithm, an improved TLD algorithm based on multi-innovation Kalman filter is proposed. The improved algorithm models the target before tracking, uses the results of TLD tracking algorithm as the current observations and then combines with the predicted values of the multi-innovation Kalman filter to optimize the tracking results. The experiments show that the improved algorithm of TLD has higher precision than the original TLD algorithm, and it is able to predict the posi- tion of the target when the target is occluded.