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

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。

Abstract:

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, Kalman prediction for target tracking. The result of KAZE comparison experiment indicates that comparing the traditional feature extraction algorithm matching algorithm KAZE has higher match ratio and better robustness. The application of Kalman prediction algorithm with KAZE in stereo visual tracking and measurement system under Visual Studio environment shows that the prediction accuracy of binocular tracking has been improved, and achieve better accuracy in the medium-long range.