Novel low-cost mobile mapping systems for forest inventories as terrestrial laser scanning alternatives

The development of devices capable of generating three-dimensional (3D) point clouds of the forest is flourishing in recent years. It is possible to generate relatively dense and accurate 3D data not only by terrestrial laser scanning but also mobile laser scanning, personal laser scanning (hand-held or in a backpack), photogrammetry, or even using smart devices with Time-of-Flight sensors. Each of the mentioned devices has their limits of usability, and different method to capture and generate 3D point clouds needs to be applied. Therefore, the objective of our experiment was to compare the performance of low-cost technologies capable of generating point clouds and their accuracy of tree detection and diameter at breast height estimation. We tested a multicamera prototype (MultiCam) for terrestrial mobile photogrammetry constructed by authors. This device is capable of capturing images from four cameras simultaneously and with exact synchronization during mobile data acquisition. Secondly, we have designed and conducted a data collection with iPad Pro 2020 using the new built-in LiDAR sensor. Then we have used mobile scanning approach applied a hand-held personal laser scanning (PLShh) using GeoSlam Horizon scanner. Moreover, we have used terrestrial laser scanning (TLS) using FARO Focus s70. With all mentioned devices, we have focused on individual tree detection and diameter at breast height measurements by cylinder-based algorithm across eight test sites with dimensions 25x25 m. Altogether, 301 trees were located on test sites, and 268 were considered for the analysis and comparisons (DBH > 7 cm). TLS provided the most accurate and reliable data. Across all test sites, we achieved the highest accuracy (rRMSE ranged from 3.7% to 6.4%) and tree detection rate (90.6-100%). When we have considered only trees with DBH higher than 20 cm, the tree detection rate was 100% across all test sites (altogether 159 trees). When the threshold of trees considered in the evaluation was changed to 10 cm and then to 20 cm (from 7 cm), the accuracy (rRMSE) and tree detection rate increased for all devices significantly. Results achieved (DBH > 7 cm) by iPad Pro were closest to TLS results. The rRMSE ranged across test sites from 8.6% to 12.9% and tree detection 64.5% to 87.5%. PLShh and MultiCam, the rRMSE ranged from 13.1% to 24.9% and 14% to 38.2%, respectively. The tree detection rate ranged from 55.6% to 75% and 57.1% to 71.9%, respectively. The time needed to conduct data collection on a test site was fastest using MultiCam (approx. 8 min) and slowest using TLS (approx. 40 min).