ALS (Airborne Lidar) accuracy: Can potential low data quality of Lidar ground points be modelled/detected based on recorded point cloud characteristics? Case study of 2016 Lidar capture over Auckland, New Zealand

Abstract: Gabriela Olekszyk ALS (Airborne Lidar) accuracy: Can potential low data quality of Lidar ground points be modelled/detected based on recorded point cloud characteristics? Case study of 2016 Lidar capture over Auckland, New Zealand. Lidar (Light Detection and Ranging) data is becoming more widely available and accessible. In many cases, it can be obtained free of charge from government agencies or local councils. In order to effectively use it in applications that require high precision, the data must be carefully studied, and sometimes verified with high precision terrestrial survey, to avoid issues introduced by potentially low point cloud accuracy. Accuracy of Lidar data is influenced by multiple factors, such as instrument position and internal errors, distance to measured surface, errors in point detection, wrong classification or complex, sloping terrain. This research focuses on analysing if recorded point characteristics, as well as some point cloud shape characteristics, show a relationship with poor data accuracy. Data used in this study was obtained for and distributed by Auckland Council in New Zealand. The available point cloud covers a large portion of Auckland and its surroundings. LAStools software has been used to manipulate the point cloud and extract various characteristics for 5m by 5m grid cells. Tested variables included: The number of present classes in a cell, the density of ground points (also after applying thinning algorithms), the height range and standard deviation of ground points, the intensity range, the average value and standard deviation, the average number of returns, the average scan angle, and the slope. Correlation analysis and multiple regression have been performed and no significant relationship was found between the tested variables and data accuracy using this research paper’s methodology. When comparing ground and low vegetation classes, some point cloud characteristics trends have been found, however, these are not suitable to aid with misclassification detection. Failure to detect meaningful relationships between recorded point cloud characteristics and accuracy or misclassification errors does not definitevely mean that there are none. Different methods could lead to more promising outcomes. Keywords: Geography, GIS, Lidar, ALS, Point cloud, Accuracy, Quality Advisor: Hongxiao Jin Master degree project 30 credits in Geographical Information Sciences, 2022 Department of Physical Geography and Ecosystem Science, Lund University Thesis nr 140