摘要
论文利用模糊推理方法提出了一种基于多特征融合的粒子滤波跟踪算法。该算法不但有效继承了传统的固定权值融合方法,并且依据模糊推理对跟踪期间信息的可靠性来输出权值的大小。根据目标模型的形状信息和颜色信息特征的观测似然函数获取各自在跟踪过程中的权值;依据模糊推理,对跟踪期间某一个变化明显而丢失目标的特征信息改变其权值,同时相应的改变另一个特征信息的权值继续无误差的来跟踪目标。与现有的经典的算法相比,提出的算法有更好的跟踪性能及较小的定位误差。实验结果表明了论文所提出算法的有效性。
This paper proposes a modified particle filter algorithm based on multi-feature fusion by using the fuzzy reasoning method.The method effectively follows fusion methods with the fixed weight,at the same time,the fuzzy inference based upon the reliability of tracking information to decide the weight.We seek the each tracking weight ratio according to the observation likelihood function of target models sharp and color feature.On the basis of fuzzy reasoning,when the feature information changed during the course of tracking,we can change its weight ratio,at the same time change another feature information weight ratio to continue tracking target without error.Compared with the existing classic algorithms,the proposed algorithm has a better tracking performance property and smaller positioning error.The experimental result shows the better effectiveness than the presented algorithm.
出处
《计算机与数字工程》
2017年第2期347-350,402,共5页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61040055)
陕西省教育厅科学研究项目(编号:2013JK1109)资助
关键词
模糊推理
特征融合
观测模型
粒子滤波
fuzzy reasoning
feature fusion
observation model
particle filter
