Sequence alignment algorithm is very mature.Smith-Waterman’s dynamic algorithm is the most efficient and basic method in 1-dimensional sequence alignment.But it is not fit for 3-dimensional structure alignment because it requires vast time and space.In this paper

we develop a dynamic algorithm to compare two protein 3-dimensional structures by using some geometry invariants under translation and rotation.Geometry invariants include many useful features of protein 3-dimensional structure.By calculating the geometry invariants

"distance" and "similarity scoring" are defined.At the end

a new algorithm based on dynamic programming is provided.Comparison with computer simulation and HOMSTRAD（HOMologous Structure Alignment Database） indicates that our approach is better in both the aligned results and alignment speed.