In order to locate weak target with sub pixel precision under strong noise condition, a sub pixel edge detection algorithm was proposed based on improved wavelet transform and Zernike moments. Firstly, wavelet transform and modulus maxima were combined to search the edge points “from coarse to fine” to achieve accurate and effective locating, then sub pixel edge points were detected by the improved Zernike moments. The method was used for dealing with images with noise in actual engineering application. The experimental result showed that the algorithm has stronger anti noise capability and higher locating accuracy, and it is a stable and reliable algorithm with better performance over that of the traditional edge detection algorithms.