Abstract: A new approach to fit line is proposed.In this method,Hough transform and least square have been combined to process experiment data and the contour of binary images.Firstly,interferential points and noise in the set of data points have been deleted by using Hough transform,meanwhile the points in the vicinity of different lines are separated; Secondly,lines have been fit by using least square.When fitting line using least square,it always encounters some problems.On the one hand,it can't overcome the interference of interferential points and noise,on the other hand,when data points distributed in the vicinity of a few lines,it is difficult to fit line based on these points.When fitting lines using Hough transform,it tends to have low precision and difficult to control line in the valid trivial.The presented method can overcome these problems.The effectiveness of the algorithm is demonstrated by the experiment.