基于均值漂移和模糊控制的主动视觉系统  

Active Vision System Based on Mean Shift and Fuzzy Control

作　　者：陈万龙[1] 彭冬亮[1] 任刚[1] 谷雨[1] 

机构地区：[1]杭州电子科技大学通信信息传输与融合技术国防重点实验室,杭州310000

出　　处：《控制工程》2017年第9期1946-1951,共6页Control Engineering of China

摘　　要：实现了一种基于PT转台的主动视觉系统。为满足实时性要求,在均值漂移(Mean Shift)踪框架内,场景光照变化会引起目标模型退化,进而导致目标失跟,针对这一情况,采用初始帧模板、快变模板、慢变模板等对模型进行线性加权更新;针对目标的快速运动而导致的失跟现象,通过引入卡尔曼预测估计目标的可能位置;以视觉跟踪器输出的像素坐标偏差及偏差变化率作为输入,以转台运动速度作为输出,设计模糊控制器,实现离线计算,在线查表的方式,实现对感兴趣目标的实时跟踪。在多种场景下对提出的改进均值漂移算法和设计的模糊控制器进行了测试,实验结果表明,系统能够稳定、快速跟踪运动目标,降低了目标丢失概率。An active vision system based on PT turnable table is realized. To meet the requirements of real-time, in the visual tracking framework of mean shift algorithm, due to the degradation of the target model caused by the change of scene illumination, which will then lead to target tracking failure, the target model is updated by using weighted templates, including the initial template, quick-change template and slow-change template. Aiming at target tracking failure caused by fast moving, Kalman prediction algorithm is introduced into the system to estimate the possible position of the target. Making the output of the vision tracker as the input of the fuzzy controller and the table motion rate as the output of it, the fuzzy controller is designed, which takes the method of off-line calculation and on-line list checking, to realize real-time tracking of the interesting target. The improved mean shift algorithm and the fuzzy controller are tested in a variety of scenarios and the experiment results show that the system can track the moving target stably and rapidly, and reduce the probability of target tracking failure.

关 键 词：主动视觉 均值漂移 卡尔曼预测 模糊控制 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]