<p><strong>Abstract.</strong> In recent years, the requirements in the industrial production, e.g., ships or planes, have been increased. In addition to high accuracy requirements with a standard deviation of 1<span class="thinspace"></span>mm, an efficient 3D object capturing is required. In terms of efficiency, kinematic laser scanning (k-TLS) has been proven its worth in recent years. It can be seen as an alternative to the well established static terrestrial laser scanning (s-TLS). However, current k-TLS based multi-sensor-systems (MSS) are not able to fulfil the high accuracy requirements. Thus, a new k-TLS based MSS and suitable processing algorithms have to be developed. In this contribution a new k-TLS based MSS will be presented. The main focus will lie on the (geo-)referencing process. Due to the high accuracy requirements, a novel procedure of external (geo-)referencing is used here. Hereby, a mobile platform, which is equipped with a profile laser scanner, will be tracked by a laser tracker. Due to the fact that the measurement frequency of the laser scanner is significantly higher than the measurement frequency of the laser tracker a direct point wise (geo-)referencing is not possible. To enable this a Kalman filter model is set up and implemented. In the prediction step each point is shifted according to the determined velocity of the platform. Because of the nonlinear motion of the platform an iterative extended Kalman filter (iEKF) is used here. Furthermore, test measurements of a panel with the k-TLS based MSS and with s-TLS were carried out. To compare the results, the 3D distances with the M3C2-algorithm between the s-TLS 3D point cloud and the k-TLS 3D point cloud are estimated. It can be noted, that the usage of a system model for the (geo-)referencing is essential. The results show that the mentioned high accuracy requirements have been achieved.</p>
Dense 3D Point Cloud Reconstruction Using a Deep Pyramid Network