With the development of unmanned aerial vehicle and digital image processing technology，surface disease recognition methods based on computer vision have been widely used in bridges and roads because of their safety and rapidity.However，due to the complexity of images，it is challenging to achieve accurate segmentation and quantitative analysis of wharf apparent diseases by current methods.This paper proposes a method for segmentation and quantitative calculation of wharf surface spalling disease regions based on zero-shot deep learning model SAM (Segment Anything Model) and image transmission transformation correction.SAM can effectively overcome the problems of concrete spalling images such as background noise，small gray difference and so on，thus being more accurate and less sensitive by noise.Further，the quantitative calculation of key geometric features of spalling diseases is realized through image processing operations such as correction，denoising and conversion.The experiments based on laboratory model and field images show that the new method has strong generalization ability and high accuracy，and can realize the accurate and rapid recognition of wharf concrete surface diseases and has a wide application prospect.