Empirical Accuracy Assessment of MMS Laser Point Clouds
The aim of study is to establish positional accuracy evaluation methodology for Mobil Mapping System (MMS) based point cloud measurement data. We will introduce a new methodology for the quality assessment of MMS point cloud data and test our approach in practical experiments. In recent years 3D measurement technologies have advanced significantly. Mobile Mapping Systems (MMS ) have become common measurement instruments for road facility management and related fields. PASCO has been utilizing on Mitsubishi Electronic Corporation (MELCO) MMS systems for road facilities census investigation since 2008. In this study, the authors used MELCO MMS-X. Its configuration consists of three GPS receivers, an IMU and an odometer, six digital cameras and four laser scanners(Figure1). In this study we establish a positional accuracy evaluation methodology for MMS based on point clouds, here derived from MMS-X Laser measurements. The methodology is generic in the sense that it can also be applied to point clouds derived by image matching algorithms from images. The general approaches of point cloud accuracy assessment can be divided into two main types. The first approach is a methodology based on the comparison of a few individual check points. This approach is not able to evaluate the accuracy of the whole point cloud. The second approach is based on the comparison of surfaces. We adopted the Least Squares 3D Surface Matching (LS3D) technique for the assessment of point cloud accuracies. This approach is an advanced methodology of co-registration of point clouds (Gruen and Akca, 2005; Akca, 2010). This method is an extension and generalization of the 2D least squares image matching into 3D space. For the purpose of experimental testing, we set up a test field in Yokohama, Japan under the consideration of good GPS receiving conditions. In 3D LS surface matching two surfaces are compared: A reference (template) surface with the actual (search) surface to be analyzed. As for template surface generation, we measured various artificial planar structures by traditional surveying methods and generate their surfaces. We selected these artificial structures under consideration of distances from the vehicle trajectory. We defined near range as 10m distance, middle range as 10-40m distance and far range as 40-80m. In this traditional survey, we constructed a geodetic reference frame work. Thus we acquired two stand-alone geodetic reference frame works at both sides of the road by GPS and traverse surveys. In order to generate several kinds of MMS point clouds with different point densities, several MMS runs with different vehicle moving speeds were executed. In our experiment, our point cloud accuracy assessment consists of relative evaluation and absolute evaluation. The relative evaluation compares overlapping surfaces generated by the MMS. The absolute evaluation compares template surfaces generated by traditional ground surveying and/or terrestrial LiDAR to surfaces generated by the MMS. We will report about the theoretical aspects of our methodology and about the results achieved in practical experiments. References:Gruen, A., Akca, D., 2005. Least squares 3D surface and curve matching. ISPRS Journal of Photogrammetry and Remote Sensing, 59 (3), 151-174. Akca, D., 2010. Co-registration of surfaces by 3D Least Squares matching. Photogrammetric Engineering and Remote Sensing, 76 (3) , 307-318.
|Laser scanning||Point Cloud||Accuracy||Mobile|