Abstract: D invariant is a kind of feature parameter that can stay consistent along with the change of pose, view point or other imaging conditions. It can be used widely in computer vision. The approaches to calculation of 2D projective transformation matrix are extended from using points to using points, lines and point line groups. So the application conditions for describing the relationship between two projective planes become free and wide. From this, the extraction of virtual elements based on various kinds of point line groups are presented, and many kinds of 3D invariant of complicated spatial structures can be neatly extracted by combining real elements and virtual elements. These invariants can be used to object recognition and interpretation. The above methods have been verified by experiments.