In-situ detection of apple fruit using a 2D LiDAR laser scanner

The development of reliable fruit detection and localization systems has been approached with remote sensing methods aimed at optimizing the orchard management to obtain high crop quality and economically improved harvesting practice. This work presents a new technique that uses a light detection and ranging (LiDAR) laser scanner to detect and localize apple fruits in the orchard. In a 1 ha apple orchard (Malus x domestica 'Gala') two trees were defoliated before harvest period. A LiDAR scanner emitting at 905 nm, with a real time kinematic global navigation satellite system to geo-reference the data and an inertial measurement unit to acquire orientation data were mounted on a tractor (0.2 km/h) to produce the 3D tree point cloud before and after defoliation. Subsequently, the apples of each tree were harvested and classified in four size classes according to height (Hmanual) and diameter (Dmanual).An intensity analysis of tree elements was performed, obtaining mean intensity values of 28.9%, 29.1%, and 44.3% for leaves, branches and trunks, and apples, respectively. These results suggested that the intensity parameter can be useful to detect apples. A four-step fruit detection algorithm was developed to localize and estimate the height (HLiDAR) and diameter (DLiDAR) of fruits. A mean detection success of 92.5% was obtained in relation to the total amount of fruits on the defoliated trees during the stages of fruit development. A mean correlation of R2 = 0.83 was obtained for Hmanual and HLiDAR, whereas a less pronounced relation was observed between DLiDAR and Dmanual (R2 = 0.62) during fruit development. The mean detection success was decreased to 70.5% when the fruit detection algorithm was applied in the foliated trees. From the experimental results, it can be concluded that LiDAR-based technology and, particularly, its intensity information has potential for remote apple detection and 3D localisation.