Rapid detection algorithms for log diameter classes based on stereo vision

Abstract

Log Diameter Classes 3D measurement with binocular stereo vision system was adopted. According to the log-end histogram feature, a region labeling method was proposed based on the maximum entropy threshold segmentation. Adopting the region labeling based on pixel labeled method of connecting area, each log could be precisely identified and counted by the system. Extraction of log edge using a Canny operator, completed stereo matching based on the epipolar line rectification, then obtained the matching 3D coordinate points. According to the quasi-circular of the log ends, selected the appropriate initial value to fit the elliptic boundary value. The smallest Euclidean distance between the boundary points and fitting points was calculated by using least squares principle, thus the best fitting ellipse and log diameter class parameters of major axis and minor axis were gotten. Experiment shows that the proposed algorithms can accurately and rapidly detect the log diameter classes.