This paper presents a stable least squares ellipses fitting algorithm based on Zernike orthogonal moments. The algorithm uses Zernike orthogonal moments for sub-pixel edge detection, and a mask of seven multiply seven was derived in the meantime. The optimal ellipse parameters were computed according to the data points extracted previously. This stable, robust and non-iterative algorithm can be easily implemented. The experiment results show that the proposed algorithm is effective in various situations.