3D fruit orchard reconstruction and geometric parameter estimation using LiDAR-based SLAM systems

Accurate knowledge of fruit tree geometry is essential for optimizing the efficiency and sustainability of agricultural operations through precision agriculture approaches. LiDAR technology enables high-resolution 3D reconstructions of fruit orchards; however, its high cost remains as a major limitation. Recent developments in low-cost solid-state LiDAR sensors offer promising opportunities for the electronic characterization of fruit trees. In this study, three LiDAR sensor systems with varying costs were evaluated in combination with SLAM (Simultaneous Localization and Mapping) techniques to obtain 3D reconstructions of an apple tree row and to estimate key geometric parameters such as tree height, tree width, and crown cross-sectional area. The systems assessed comprised a commercial backpack mobile terrestrial laser scanner (used as ground-truth), a digital spinning LiDAR sensor, and a low-cost hybrid solid-state LiDAR sensor. Results showed that while all systems provided consistent tree height estimations, significant differences emerged in the measurements of tree width and cross-sectional area. The digital spinning LiDAR, with its high point rate and structured mechanical rotation, produced more accurate and repeatable measurements, particularly in forward scanning trajectories. In contrast, the low-cost sensor’s non-repetitive scanning pattern and lower point density led to increased errors, especially in round-trip trajectories due to accumulated SLAM drift. These findings underscore the importance of both sensor specifications and scanning strategies in achieving reliable SLAM-based 3D reconstructions in orchard environments. Striking a balance between accuracy and cost is therefore crucial to support the adoption of these technologies within the fruit growing sector.