The association graph unifies the geometry constraints for pairwise candidates, in which the node represents the point or line candidate and the edge represents the geometry constraint. Compared with other graph based algorithm (Beder, 2004, McAuley et al., 2010, Yang and Liu, 2017), the association graph in GLSM is unique for two aspects: (1) the geometry constraint is manly based on the local coplanar assumption, and the edge between line and point candidate is able to be established. (2) The