Size and Pose Identification of Bolts Based on Monocular Vision

Abstract

Machine vision has extensively developed in automatic image recognition research, and the automatic identification of bolts can improve the degree of automation greatly for bolts operation. This paper provides a method to extract the outline and the geometrical shape of the bolt thereby to identify the target bolt and locate the position according to the shape and the size of a kind of common six-angle bolt. A dynamic threshold segmentation method is used to extract the bolt geometry information and set the target area as region of interests (ROI). By the sub pixel boundary calculation, Ramer algorithm for segmentation of contours, and Tukey weight function with five iterations for linear fitting, the intersection features which are the bolt corners can be obtained. Through calibration, the relative position of bolt can be located and the size of bolt can be acquired.