Efficient point cloud corrections for mobile monitoring applications using road/rail-side infrastructure

Abstract

Light Detection and Ranging (LiDAR) systems are known to capture high density and accuracy data much more efficiently than other surveying methods. Therefore they are used for many applications, e.g. mobile mapping and surveying, 3D modelling, hazard detection, etc. However, while the accuracy of the laser measurements is very high, the accuracy of the resulting 3D point cloud is greatly affected by the geo-referencing accuracy. This is especially problematic for mobile laser scanning systems, where the LiDAR is installed on a moving platform, e.g. a vehicle, and the point cloud is geo-referenced by the data provided by a navigation system. Owing to the complexity of the surrounding environments and external conditions, the accuracy of the navigation system varies and thereby changes the quality of the point cloud. Conventional methods for correcting the point cloud accuracy either rely heavily on manual work or semi-automatic registration methods. While they can provide geo-referencing under different conditions, each has their own problems. This paper presents a semi-automated geo-referencing trajectory correction method by extracting features from the pre-processed point cloud and integrating this information to reprocess the navigation trajectory which is then able to produce better quality point clouds. The method deals with the changing errors within a point cloud dataset, and reducing the trajectory error from metre level to decimetre level, improving the accuracy by at least 56%. The accuracy of the regenerated point cloud then becomes suitable for many accuracy-demanding monitoring and change detection applications.