Identification of Log Characteristics in Computed Tomography Images of Black Spruce (Picea Mariana) Logs by Means of Maximum Likelihood Classifier

Abstract

As a nondestructive technique, computed tomography (CT) has been studied to acquire internal log images in the past twenty years. The produced CT images are true cross-sectional images of logs. The main objective of this study is to investigate the feasibility of identifying internal log characteristics in black spruce CT images using the maximum likelihood classifier. The log characteristics to be identified include heartwood, sapwood, bark and knots in black spruce. Twenty CT images were randomly sampled from one black spruce log to develop the classifier and additional twenty images were obtained from another log for validation. The diagnostic values of image features including gray level value, distance and textural features, in distinguishing the log characteristics were assessed first based on the class separability analysis. Useful image features were then employed to develop the classifier. The achieved classification overall accuracies were 80.9% and 71.0% for the training images and the validation images, respectively. This study indicates that the maximum likelihood classifier may be applicable to identify the internal log characteristics in CT images of black spruce. The results also reveal that the separability of one log characteristic from the other log characteristics in CT images is mainly related to its physical properties including wood density and moisture content.