基于非线性扩散滤波的双目视觉跟踪与测量研究实现  

Stereo Vision Tracking and Measurement System based on KAZE feature

作　　者：于菲[1] 战兴群[1] 刘铭[1] 

机构地区：[1]上海交通大学航空航天学院,上海200240

出　　处：《计算机测量与控制》2016年第2期30-33,共4页Computer Measurement &Control

基　　金：国家高技术研究发展计划(863计划)(2014AA123103)

摘　　要：针对双目视觉跟踪与测量技术的研究主要集中在缩短时间,精确性问题考虑不够充分;结合KAZE特征提取算法,利用卡尔曼预测对目标进行跟踪,提出一种基于图像匹配的双目视觉跟踪与测量算法;将KAZE特征提取算法和SIFT、SURF等算法进行仿真对比,结果表明KAZE算法具有更加出色的匹配率,而跟踪速度相当;运用KAZE算法与卡尔曼预测跟踪技术实现了基于Visual Studio环境下的双目视觉测量与跟踪系统,测试结果表明,该系统可进行有效的高精度测量。The intention of this research is achieve an accurate stereo visual tracking and measurement system. The emphasis of this application lies in the KAZE feature, a novel image matching algorithm, Nalman prediction for target tracking. The result of NAZI; comparison experiment indi- cates that comparing the traditional feature extraction algorithm matching algorithm KAZE has higher maleh ratio and better robustness. The applica- tion of Kalman prediction algorithm with KAZE in stereo visual tracking and measurement system under Visual Studio environment shows that the pre- diction accuracy of binocular tracking has been improved, and achieve better accuracy in the medium long range.

关 键 词：双目视觉 图像匹配 非线性扩散滤波 视觉跟踪 卡尔曼预测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]