Tree Defect Segmentation Using Geometric Features and CNN

Estimating the quality of standing trees or roundwood after felling is a crucial step in forest production trading. The on-going revolution sector resulting from use 3D sensors can also contribute to this step. Among them terrestrial lidar scanning reference descriptive method offering possibility segment defects. In paper, we propose new reproducible allowing automatically It based on construction relief map inspired previous strategy and combining with convolutional neural network improve segmentation quality. proposed outperforms results source code publicly available an online demonstration test defect detection without any software installation.