Deep learning for tree trunk imaging via tree radar

Abstract

The wood business and determining the health of living trees both benefit from tree trunk examinations. Non-destructive testing (NDT) approaches for diagnosing areas in living trees are becoming increasingly common. Ground Penetrating Radar (GPR) is widely recognized as a particularly effective instrument for monitoring tree trunks among the current NDT inspection methods. Low-resolution photos, on the other hand, can make it difficult to spot flaws in the wood.

This paper is the report for the final year project entitled ‘Deep learning for tree trunk imaging via tree radar’. The purpose of this report is to document the project’s progression and accomplishments, as well as any problems that may have arisen and ongoing work. This report is 44 pages long, excluding the cover page, abstract, acknowledgement, list of figures and tables, bibliography and appendix.

The main aim of this project is to use deep learning techniques for radar imaging of tree trunks with defects. Firstly, a large set of 2D radar images of the tree trunks with defects and noise will be generated. Convolutional Neural Networks will then be used for denoising these images, getting the images of the tree trunks as well as the defects.