Merging Point Clouds from Heterogeneous 3D Scanners for Fast Update of an Earthwork Site Model

In earthwork sites, its terrain shapes are changing continuously as the work proceeds. If someone wants to grab three-dimensional geometry data of such sites, currently available technologies such as 3D laser scanners are unable to collect reasonably precise data in pseudo-real time (i.e., 1-10 scans per second). To tackle the problem, the authors took ‘dual scanners’ approach, where low-speed (yet highly-precise) laser scanners are used for initial scan of the work site, then augment the initial data using the 3D geometry data collected from excavator-mounted stereovision. For this approach, the authors used ICP (Iterative Closest-Point)-based algorithms for merging the stereovision data with the laser scan data, where similarities of both data are limited due to changed site shape.