Abstract: For extracting the feature lines in welding seam images,the improved least-square method based on linearity is proposed.This improved least-square method defines three parameters:linearity L;the limit of distance between points and the fitted lines R,the number of the data points eliminated at one time N.Taking L as principle,with R and N,the data points of the centerline are divided into several point sets by using the least-square method iteratively.The feature line is included in the point set which contains most of the points,and then,least-square method is applied to fit this point set until the linearity of the fitted line is not less than the expected one.This fitted line is taken as the feature line.Finally,the influence of three parameters on the fitting lines is discussed.The results indicate that this method can efficiently extract the feature lines in V-groove welding seam image and meet the requirement of image processing methods applied in seam tracking.