An inversion algorithm for  aerosol optical depth based on different vegetation covers is introduced. This algorithm is designed to estimate the precise Aerosol Optical Depth (AOD) by using the visible to near-infrared spectral bands (400-900nm) information. We describe the principle of this method and outline the main steps in this correction process. Simulated data and remotely sensed images are used to validate this algorithm. The fitted  coefficient R2 equals to 0.981 8 and the error of root mean square(RMS) equals to 0.055, which shows that the method has a good performance. Moreover, a CBERS(China-Brazil Earth Resource Satellite) CCD image is selected to validate this method. The surface reflectance after atmospheric correction is much closer to the real spectrum. The method is conducive to the quantitative application of CCD images.