Orchard tree digitization for structural-geometrical modeling

摘要

This paper presents a system for digitizing the geometries and representing the topologies of orchard trees. Digitization is performed manually using commercial electromagnetic sensors arranged in a custom built frame. Digitization is restricted to rigid, thick branches (>2.5 cm) that present obstacles to machinery interacting with trees; robotic harvester design is the focus application. Manual digitization is slow, but it can provide accurate tree geometries. Each branch segment that has small enough curvature to be considered linear is approximated by a conical frustum. Hence, the entire tree is represented geometrically by a set of connected frustums. Experimental data from ten Bartlett pear trees in a commercial orchard were acquired and the corresponding tree geometries and topologies were successfully encoded with RMS accuracy better than 1.0 cm.