Line Segment-based Approach for Accuracy Assessment of MLS point clouds in Urban Areas

International audienceThis paper presents an accuracy assessment of Mobile Laser Scanning (MLS) point clouds and the initial results of practical experiments carried out for a custom-built mobile system. To minimise the difficulties of identification and precise measurement of control points, an approach based on line segments is proposed. The main aim of this technique is to compare a series of 3D line segments extracted from MLS data with those from a reference data set. The lines represent edges found as intersections between the principal planes previously modelled within the MLS point cloud. In order to calculate the modified Hausdorff distance, a correspondence-finding algorithm automatically locates similar line segments inside two different data sets: Test and Model. The results achieved confirm that the proposed method appears to be effective for accuracy assessment, especially when evaluating low density point clouds. The experiment site for data acquisition was located in an urban area rich in tall buildings and narrow streets. An estimation of the precision of the mobile system was made using repeated data collection passes, indicating a measurement error value of about 0.3 m. The absolute accuracy was evaluated by comparison with a reference point cloud collected by a static laser scanning survey. The assessment indicates an average difference of less than 0.7 m along two-way street with relatively unobstructed view of the sky