Zernike moments are complex moments with the orthogonal Zernike polynomials as kernel function, compared with other moments; Zernike moments have greater advantages in image rotation and low noise sensitivity. Because of the Zernike moments have image rotation invariance, and can construct arbitrary high order moments, it can be used for target recognition. In this paper, the Zernike moment algorithm is improved, which makes it having scale invariance in the processing of digital image. At last, an application of the improved Zernike moments in image recognition is given.