基于单目视频的伪双目动态测距方法

Pseudo binocular dynamic ranging method based on monocular video

传统的双目测距是通过待测物体在双目相机中成像的视差来计算物体在三维空间中的坐标信息,需要结合理想的双目测距模型来完成。然而在实际的工程应用中所使用的摄像机往往是单目摄像相机,无法满足双目测距要求。针对以上问题提出一种基于单目视频的伪双目动态测距方法,针对在预定轨道上移动的摄像机拍摄的视频,对目标物体进行测距。对所使用的单目摄相机进行标定,得到摄相机的焦距、主点坐标等内部参数。在摄像机拍摄的视频中截取不同位置时的两帧图像,通过轨道参数计算出在不同位置时摄像机的位姿关系,运用归一化相关的模板匹配算法,得到最佳的像素匹配点,结合伪双目测距模型计算待测物体的距离,并与传统双目测距以及单目测距进行了比对。实验结果表明,当摄相机距待测物体5 m以内时,在摄相机视野范围内测量物体,误差能够维持在4%以内。

The traditional binocular distance measurement is to calculate the coordinate information of the object in the three-dimensional space through the parallax of the object to be measured in the binocular camera, which needs to be combined with the ideal binocular distance measurement model. However, the cameras used in actual engineering applications are often monocular cameras, which cannot meet the requirements of binocular distance measurement. Aiming at the above problems, a pseudo-binocular dynamic ranging method based on monocular video is proposed, and the target object is ranging from the video taken by a camera moving on a predetermined track. Calibrate the monocular camera used to obtain the camera′s focal length, principal point coordinates and other internal parameters. Intercept two frames of images at different positions in the video taken by the camera, calculate the pose relationship of the camera at different positions through the track parameters, and use the normalized template matching algorithm to obtain the best pixel matching point, combined with pseudo. The binocular ranging model calculates the distance of the object to be measured, and compares it with traditional binocular ranging and monocular ranging. The experimental results show that when the camera is within 5 m of the object to be measured, the error can be maintained within 4% when measuring objects within the camera′s field of view.