基于立体视觉的被动测距技术研究

为了实现基于立体视觉的被动测距技术,该文搭建了一套双目成像测距系统实验平台,研究了基于双目立体视觉的被动成像处理算法;设计了可实现多目标的实时跟踪测距的双目成像测距应用软件,对提出的一种基于SIFT搜索窗口大小随着目标尺寸变化自适应调整的跟踪算法进行了验证.测试结果表明该软件在跟踪、测距方面表现出色,在1~2个目标时具有较好的实时性.通过平面标定法进行准确的摄像机标定,采用SIFT特征匹配算法进行立体匹配,测试结果表明在取octave层数为4时,匹配准确度一般可在0.05个像素范围内.实际外场测试实验表明,该算法对一定距离范围内的目标物体进行测量具有较好的准确性.

On Passive Ranging System Based on Binocular Stereo Imaging

In order to realize the passive ranging technology based on stereo vision,in this paper a binocular ranging system experimental platform has been established,the passive imaging processing algorithm based on binocular stereo vision studied,and the binocular ranging application software designed,which can realize the real-time tracking of multi targets ranging.Finally,a SIFT search window size with the tar-get tracking algorithm based on adaptive adjustment of the size change has been verified.The test results show that the software has a good performance in tracking,ranging,and has a better real-time perform-ance in the 1-2 target.For camera calibration by plane calibration method and use the SIFT feature matc-hing algorithm for stereo matching,in the octave level is 4,matching accuracy generally can reach 0.05 pixel range.Field test experiment show that measurement algorithm has a good accuracy for the target ob-ject within a certain distance.