A Novel Linear Edge Fitting Method with Short Edge Correction

Abstract

In this paper, a novel linear fitting method for micro vision is proposed, which can be utilized to accurately estimate the pose of components in micro-assembly. After obtaining the edge information through Canny edge detection, an improved least square method is put forward to fit the linear edge of small components, where a threshold is utilized to reduce the influence of outliers in point set. In order to improve the estimation accuracy, the fitted linear edge is iteratively corrected through the information of detected broken lines. Compared with conventional linear edge fitting methods, the experimental results indicated that proposed method could effectively reduce the confidence interval of the fitted line, which would be helpful to improve the accuracy of pose estimation in micro-assembly.