鲁棒的交互式多模粒子滤波跟踪器概率计算方法

发布时间：2018-06-26 06:23

本文选题：粒子滤波 + 跟踪器概率　；参考：《光电子·激光》2017年06期

【摘要】：针对传统交互式多模粒子滤波(IMMPF)跟踪器概率计算方法在复杂环境下鲁棒性不足的缺陷,提出一种新型的基于联合似然函数模型的子跟踪器概率计算方法。首先,计算基于跟踪结果与当前外观模型的巴氏距离作为瞬时似然函数,度量目标外观的剧烈变化;其次,利用l2范数规则化最小二乘算法构建目标的重构外观模型,将其与跟踪结果的误差指数函数作为平稳似然函数,度量目标的缓慢变化;然后,基于加权求和策略得到跟踪器基于多种特征的联合似然函数;最后,将建立的联合似然函数结合上一帧的先验状态交互概率完成子跟踪器概率的更新。对复杂环境下跟踪器性能的在线评估对比结果验证了联合似然函数模型能有效评估跟踪器因不同干扰因素导致的性能变化,将其应用于子跟踪器概率的计算能获得比主流算法更好的鲁棒性。
[Abstract]:Aiming at the shortcoming of the traditional interactive multi-mode particle filter (IMMPF) tracker probability calculation method which is not robust in complex environment, a new sub-tracker probability calculation method based on the joint likelihood function model is proposed. Firstly, the pasteurian distance based on the tracking result and the current appearance model is calculated as the instantaneous likelihood function to measure the drastic changes of the target appearance. Secondly, the reconstruction appearance model of the target is constructed by using the L2-norm regularized least squares algorithm. The error exponent function of tracking result is regarded as stationary likelihood function to measure the slow change of target. Then, the joint likelihood function based on multiple features is obtained based on weighted summation strategy. The proposed joint likelihood function is combined with the prior state interaction probability of the previous frame to complete the update of the sub-tracker probability. The online performance evaluation of tracker in complex environment shows that the joint likelihood function model can effectively evaluate the performance changes caused by different interference factors. By applying it to the calculation of the probability of subtracker, the robustness of the algorithm is better than that of the mainstream algorithm.
【作者单位】：电子工程学院红外与低温等离子体安徽省重点实验室脉冲功率激光技术国家重点实验室;电子工程学院训练部教务处;
【基金】：国家自然科学基金(61405248,61503394) 安徽省自然科学基金(1408085QF131,1508085QF121) 安徽高校自然科学重大研究项目(KJ2015ZD14)资助项目
【分类号】：TN713;TP391.41

【相似文献】