Comparison of segmentation methods on images of energy plants obtained by UAVs

摘要: Bioenergy crops are widely used as a form of renewable energy. The special it is the energy forestry that includes short rotation coppice plantations in which fast-growing species tree or woody shrub grown (e.g. willow, poplar). It important to accurately estimate yield before harvest order maximize profit and reduce costs production [1]. accurate prediction forest biomass volume can be for evaluation plant breeding efficiency well. Since forestries often contain different trees estimating their essential find segments containing same image. In recent times, use Unmanned Areal Vehicles (UAV) became more popular precision agriculture [2], [3]. We investigated applicability low cost UAV field agricultural image segmentation first stage estimation [4]. This paper compares 3D reconstruction methods performed on aerial photographs plants focuses applied plantation. this, several available algorithms were evaluated, such eCognition other free software tools like Orfeo Toolbox. Matlab implementations also Mean-shift [5], Statistical region merging [6] Link-based clustering ensembles [7]. best accuracy according Dice similarity coefficient was reached by supervised process Multi-resolution 74.85%. obtained an automated method, segmentation, 66.12 %.