A Multi-camera System for Efficient Point Cloud Recording in Close Range Applications

High resolution point cloud recording in close range applications is a demanding task for large objects. Current methods like handheld laser scanners are costly and require long time for acquisition. Furthermore, the recording has to be done very carefully to enable a high accuracy over the whole object. We present a flexible and compact image based solution for efficient point cloud recording in close range applications. The sensor consists of off-the-shelf industry cameras and a Microsoft Kinect which is only used for the projection of additional texture. The software uses a multi-stereo dense image matching method for the derivation of high resolution point clouds. Up to 3.5 million points can be acquired by a single shot with sub-mm accuracy and resolution at a working distance of 70cm to the object. Structure and Motion reconstruction methods are used for the registration of several shots.