Abstract: After having analyzed the disadvantage of existing automatic filtering algorithm, which means deleting the micro morphology or obstacle caused by devilishly pursuing the smoothing effect of artificial filtering method, aims at Ping as the processing cell of multi-beam bathymetric survey data, and by means of rolling circle transform's characteristic of identifying and analyzing the concave (convex) part quantitatively under specific precision, the paper details an algorithm for filtering noises in multi-beam data based on a rolling circle transform, and details the keystone and solution steps of the model. At last, in a VC++ environment, some experiments were done to validate the algorithm's validity. The experiments show that the algorithm can filter noises in multi-beam data; maintaining marine topography completeness and enhancing the efficiency in gross error detection.