Nondestructive Evaluation of Hardwood Logs Using Automated Interpretation of CT Images

Abstract

Computed tomography (CT) imaging is being used to examine the internal structure of hardwood logs. The following steps are used to automatically interpret CT images: (1) preprocessing to remove unwanted portions of the image, e.g., annual ring structure, (2) image-by-image segmentation to produce relatively homogeneous image areas, (3) volume growing to create volumes from contiguous areas on consecutive slices, and (4) object recognition to label volumes based on their CT statistics, volume, shape, and location. Experiments were conducted to test the recognition of internal log structures using several red oak and yellow poplar logs. A small set of CT slices were selected and used as training data to calculate feature distributions. Threshold values of those distributions determine distinguishing ranges of values for the defect labeling descriptors. Recognition rules were then applied to individual candidate volumes to assign confidence values to different defect classes. Recognition results are encouraging, but the software must be tested and validated on other species and on logs of differing condition, e.g., moisture content. The ability to improve current capabilities can benefit from an expanded database of CT images and their ground truth records.

Citation

Schmoldt, Daniel L.; Zhu, Dongping; Conners, Richard W. 1993. Nondestructive Evaluation of Hardwood Logs Using Automated Interpretation of CT Images. Proceedings, Review of Progress in Quantitative Nondestructive Evaluation. Issue 12: 2257-2264.