DEFECT DETECTION OF WOOD USING CONNECTED-COMPONENT LABELING

摘要

Based on industry automation requirement, we propose an image processing algorithm to detect the defective wood. RGB image is converted to binary and then feature extraction is performed by using the number of connected-component results from labeling. We use thresholding as the classification tool. The proposed algorithm reaches 96.34% accuracy. This performance is influenced by the way to convert into binary image.