Development and performance evaluation of low-cost UAV-lidar systems for forestry applications

Abstract:

Accurate and repeated forest inventory is critical for forest resource management. Recently, unmanned aerial vehicle (UAV) light detection and ranging (lidar) systems have been proven to be an effective and accurate tool for forest inventory, but the high price of UAV lidar systems has been one of the major obstacles preventing it from large-scale applications. Here, we developed a very low-cost UAV lidar system by integrating a DJI Livox Mid40 scanner ($599 in price), and evaluated its performance in both coniferous and deciduous forests with various tree densities by comparing with four high-end UAV lidar systems (equipped with RIEGL VUX-1 UAV scanner, RIEGL miniVUX-1 UAV scanner, HESAI Pandar40 scanner, and Velodyne Puck LITE scanner respectively). In each study site, all UAV lidar systems were flown repeatedly at a flight height from 50 m to 250 m with an interval of 50 m, and three flight speeds (4m/s, 6m/s, and 8m/s) were tested at each flight height.

Results show that the Livox system can obtain point cloud data with similar point densities and ground point densities in all study sites as the other four high-end UAV lidar systems, indicating the Livox system can be used to depict undercanopy terrain information effectively. However, the Livox system may miss points in the low-middle height strata due to the low sensitivity to low-reflectance targets. Moreover, the Livox system can estimate individual tree height, canopy height model (CHM), canopy cover, gap fraction, and leaf area index with similar accuracies as the other four systems in all coniferous studies sites. In broadleaved study sites, the performance of all five UAV lidar systems is worse than in coniferous sites. Among them, the RIEGL VUX-1 and REIGL miniVUX-1 systems have the highest accuracy in estimating all forest inventory attributes, followed by the HESAI Pandar40 system, the Velodyne Puck LITE system, and the Livox system. The Livox system tends underestimate height-related attributes (e.g., individual tree height and CHM). Comparing to flight speed, flight height has a stronger influence on the collected data quality, and the optimal flight height for each system is about half of the nominal maximum measuring range. We believe this study can provide guidance for the selection and future development of low-cost laser scanners for forestry applications.