This paper is focused on the acquisition of the edges of metal rectangle in the field of aluminum industry with high accuracy. The image of the metal rectangle has the reflection interference, which is not easy to get rid of in image processing. In this paper, the raw edge is acquired by gray level difference. Then, the fine edge is acquired by abandoning the noisy data. In the end, Least Square Method is used to fit the edge. According to the four edges, we get the center of the rectangle. The experiment validates the effectiveness of the algorithm. Compared to the Hough Transform, the method for edge extraction is simple and useful, and it is robust to the light interference to some extent.