Development of 3-D Range Imaging System to Scan Peach Branches for Selective Robotic Blossom Thinning

Abstract

A laser-based 3D range imaging system was proposed to generate the absolute 3D coordinates and morphology of peach trees (Prunus persica). A horizontal plane of laser light sweeps vertically through the branches, as a video camera mounted above the tree canopy captures the movement. Through a dynamical scanning approach, the background scene is automatically removed, leaving only the essential data of the peach tree branches. Images produced by the camera are then processed in real time, and absolute 3-D coordinates of the branches and blossoms are calculated using the intrinsic geometry of the laser reflections. An experimental unit was assembled indoors, and the distances from features to the camera were produced using test models. This process produces a grayscale image of the tree canopy, with pixel values equal to the distances from the camera in centimeters. Testing for this system produced accuracy results of max error emax = 1.2 cm. Error in all three directions was calculated: X (horizontally to the side) average error was eXave = 0.5 cm with standard deviation of sX = 0.2 cm; Y (vertically up from the ground) maximum error was eYmax = 1.2 cm; and Z (horizontally away from the camera) average error was eZave = 1.2 cm with standard deviation sZ = 1.1 cm. The output representative distance image is meant to guide a robotic device in the selective thinning of blossoms. This automatic selective thinning device is intended to provide an alternative to labor-intensive, manual thinning of peach trees.