基于ASODVS的全景相机运动估计及管网3D重构技术

针对地下管网的空间三维测量、三维形貌重构难等问题,提出了一种基于主动式全方位视觉传感器(ASODVS)的相机运动估计及管网3D重构解决方案。通过携带有ASODVS的管道机器人进入管道内部,实时获取管道内壁纹理全景图像和全景激光扫描图像;首先对全景激光扫描图像处理解析出投射在管道内壁上的激光中心点,通过计算得到管道横截面的点云数据;另一方面,对全景纹理图像进行处理,首先利用快速鲁棒性特征(SURF)算法快速提取特征点并进行匹配,然后采用随机抽样一致性(RANSAC)算法去除误匹配点,接着利用全景相机的极几何原理估计相机运动位姿,并利用光束法平差(BA)进行优化,最后利用相机运动位姿将相机坐标系下的点云坐标实时转换到世界坐标系下,完成对地下管网的三维重构。实验结果表明,所提出的方案能够精确估计相机运动位姿,实时对管道内部进行三维重构,实现了管道检测机器人边行走、边采集数据、边检测分析处理、边三维建模的设计目标。

Motion estimation of panoramic camera and 3D reconstruction of pipe network based on active stereo omni directional vision sensor

Aiming at the 3D measurement and 3D reconstruction of underground pipeline network, a scheme for camera motion estimation and 3D reconstruction of pipeline network based on active stereo omni direction vision sensor (ASODVS) is presented in this work. A pipe robot equipped with ASODVS travels along the pipeline, obtaining the in pipe panoramic texture images and laser scanning images in real time. The central point of the lasers projected on the inner wall is analyzed and the point cloud data of the pipe cross section are calculated. Then the panoramic texture images are processed in the following steps. First, the feature points are extracted and matched adopting the SURF algorithm; Second, the wrong matched points are removed using RANSAC method; Third, the motion poses are estimated utilizing the polar geometry principle and optimized with the BA approach; Finally, the point cloud data are converted from the camera coordinate system to the world coordinate system with the motion poses of the camera, completing the 3D reconstruction of the underground pipe network. Experimental results show that the proposed system is capable to estimate the motion poses of the panoramic camera precisely and achieve a real time 3D reconstruction of the inner pipeline, realizing the synchronization of walking, data acquisition, processing and analysis, 3D modeling for the pipeline inspection robot.