Abstract: Constellation recognition is used to identify known constellations in astronomical images and is an essential foundation for astronomical observations. Existing constellation recognition algorithms focus on using ephemerides but lack targeted research using only image information. A constellation recognition method based on triangulation and similarity constraints was proposed in this paper. This method used Delaunay triangulation to construct triangles, matching and identifying the triangles according to the similarity of internal angles, as well as realizing the preliminary screening according to the matching star points. The combination of the determined point and the test point was introduced to realize the selection of the target constellation. Experimental results show that the method proposed has an average identification accuracy and matching accuracy of 87.9% and 76.1%, respectively, in a relatively short recognition time, which can effectively improve the recognition success rate of the target constellation.