Sohn, GunhoKo, ConnieLee, Hyungju2023-03-282023-03-282022-12-092023-03-28http://hdl.handle.net/10315/41046There was an undeniable success of Deep Learning networks for visual data analytics such as object detection and segmentation in recent years, while the adaptation to tree detection has been rare. In this paper, we pursue to achieve individual tree identification, defined as a detection of an individual tree as each object, with deep convolutional neural networks to create and update tree inventories using LiDAR information. The first objective was to provide a suitable dataset that can be used to test such networks and to create a module that attempts to increase the 3D object detection algorithms' detection accuracy. This novel dataset was created by fusing LiDAR data gathered by Teledyne Optech with field data collected by York University. The second was to develop an appropriate accuracy increasing volumetric module. For this module, the learnable weights concept was introduced, which enable to increase detection precision of the object detection algorithm.Author owns copyright, except where explicitly noted. Please contact the author directly with licensing requests.Artificial intelligenceComputer scienceUrban forestryElectronic Thesis or Dissertation2023-03-28Machine learningDeep learningObject detection3D object detectionComputer visionLiDARPoint cloudLoss functionVolumetric moduleTree detection