Enhancing LiDAR Data Integration From Mobile and Airborne Laser Scanning Systems

By Tamás Radnóti

Abstract
The thesis research improved the integration of mobile laser scanning (MLS) and airborne laser scanning (ALS) systems’ dataset for forest inventory applications, through the development of a tree point clouds matching algorithm between the two datasets based on specific tree characteristics. In the following, the most important findings are summarized and the research questions are answered. The results from this research proved that tree point clouds can be matched between two different LiDAR systems based on tree characteristics, however, there are differences between the MLS and ALS systems’ accuracy in capturing features. MLS systems are particularly useful for capturing different below canopy features such as DBH with low mean error (± 0.02m), while ALS systems are more accurate at measuring canopy related characteristics like CW and CV due to their scanning perspective. Both systems are capable of measuring tree height with similar mean errors (± 0.36m; ± 0.35m), however ALS provides better accuracy for taller trees. We investigated the capabilities of each system, and the following tree characteristics were found to be measurable by both: TH, DBH, CBH, CW, CV, 3D alpha volume. On the other hand, MLS systems for CPA, CA and CD no papers were identified given the limitations of the literature review of this research. The largest MAD value differences occurred in the estimation of canopy volume (128.679), 3D alpha volume (88.758), and crown base height (15.307). These findings were aligned with peer papers as these characteristics can not be retrieved with the same accuracy in both systems. Meanwhile, characteristics such as DBH, CW and CD resulted in high accuracy between the two systems, thus proving a strong correlation and reliability. Improvement of the matching consistency between the two systems was evaluated by combining multiple tree characteristics among the most consistent ones. The newly developed algorithm was able to identify proposed matches with different combinations of characteristics, with different results. Overall, five different combinations were taken into account, this can be seen in Table 6. The best combination of tree characteristics turned out to be the CW and CD as it resulted in an overall 45.28% of matching consistency. Therefore, the combination of tree characteristics indeed can improve the matching consistency, but the combination may differ as the weight of the characteristics is not the same. The newly developed algorithm achieved good alignment consistency with the validation dataset, showcasing its potential to identify individual tree point clouds retrieved from different LiDAR scanning systems.