摘要
为了提高粒子滤波跟踪算法的效率,针对图像的特点对粒子滤波算法进行改进。首先针对无人机跟踪过程中,目标始终在图像中心位置附近晃动的特点,结合高斯分布对粒子初始化过程进行改进,保证粒子以较大概率集中在图像中心位置附近;然后针对无人机姿态调整过程中,目标运动朝向图像中心位置的特点,采用高斯加权后的权值对粒子重采样进行改进,保证图像中心区域的粒子有更大概率被保留。最后通过实验对改进算法进行分析和评价,验证了改进算法的有效性。
To improve the efficiency, the paper improved the particle filter tracking algorithm according to the image's features. At first,according to the feature that the object is always rocking near the image's center during the process of object tracking for UAV, we improved the particle initialization process to make the particles distributed near the center of the image. Then according to the feature that the object moves towards the center of the image during the process of UAV's attitude adjustment, we improved the particle re - sampling to make the particles near the center saved more possibly. At last, through the experiment, we analyzed the algorithm and verified the effectiveness of the improved algorithm.
出处
《计算机仿真》
北大核心
2017年第2期84-87,共4页
Computer Simulation
基金
国家自然科学基金项目(61172125
61132007)
国家自然科学基金-民航基金联合资助(U1533132)
关键词
粒子滤波
无人机
粒子初始化
粒子更新
粒子重采样
Particle filter
Unmanned aerial vehicle
Particle initialization
Particle update
Particle resampling