CNN-based Method for Segmenting Tree Bark Surface Singularites

Abstract

The analysis of trunk shape and, in particular, the geometric structures on the bark surface are of main interest for different applications linked to the wood industry or biological studies. Bark singularities are often external records of the history of the development of internal elements. The actors of the forest sector grade the trees by considering these singularities through standards. In this paper, we propose a method using terrestrial LiDAR data to automatically segment singularities on tree surfaces. It is based on the construction of a relief map combined with a convolutional neural network. The algorithms and the source code are available with an online demonstration allowing to test the defect detection without any software installation. Source CodeThe reviewed source code and documentation associated to the proposed algorithms are available from the web page of this article 1 . The correspondences between algorithms and source codes are pointed out, and compilation with usage instructions are included in the README.md file of the archive. Supplementary MaterialThe reference dataset, to be used for further comparisons, is provided with the article and peerreviewed. The (non peer-reviewed) instructions to generate the learned model are provided as supplementary material on the associated Github repository 2 .