Abstract: Seabed image matching is a challenge task due to image quality degradation caused by complex underwater imaging environment. In this paper, a new matching method by incorporating prior knowledge is proposed which is different from classical data-only based methods. Firstly, according to the characteristics of seabed imaging, blue channel image is selected as the original image based on the fact that blue light attenuation is minimal. By further applying appropriate image enhancement and filtering processing the quality of original image is effectively improved. Secondly, the orientation for rotation invariant descriptor is determined based on the attitude data provided by the AUV sensor, thus the accuracy of the feature descriptor is more accurate and reliable. Finally, the limited search range of corresponding feature is estimated by using the positioning and altitude data provided by the AUV sensor, which reduces the search space for feature matching. The experiment results conducted on more than 15 000 real images demonstrate that the proposed method is effective in increasing the number of feature matching and improving the percentage of correct matches. Especially for the images with low texture and repetitive texture which are more difficult to match, the method is more effective than classical ones; the percentage of correct matches is increased from 16.20% to 35.84% statistically.