Detection of Trimmed and Occluded Branches on Harvested Tree Stems using Texture Analysis
Authors
Xin Tian
New Zealand Forest Research Institute, Rotorua, New Zealand
Glen E. Murphy
New Zealand Forest Research Institute, Rotorua, New Zealand
Abstract
This paper describes a prototype computer vision system based on texture analysis that can automatically locate and identify certain classes of defects on freshly harvested tree stems (with bark present) by using digital camera imagery. This system includes two modules: (1) a feature extraction or "computing" module for estimating the oriented texture field from the raw image of the surface of a log; and (2) a scene analysis and detection module for analysing the oriented texture field. Basic principles of the system based on the gradient of the Gaussian filter have been given.
The sample digital images processed in this study are of sections of freshly harvested Radiata pine tree stems with bark still on them. The feasibility of the computer vision system has been very well demonstrated by initial experimental results.
This study has also resulted in a new taxonomy for texture description and identification in the field of forestry. Texture analysis technology has been found to be a potentially powerful approach to determine wood properties, timber and log quality.
The system developed in this study can be widely applied into forest industry activities and research.
