Spatial styles capturing using genetic algorithms in sketch understanding

Many sketches convey information through the relationships between graphics, rather than the internal structure of the graphics themselves, and the spatial context of a sketch should help people to understand freehand sketches. We exploit a method of the spatial styles capturing based on genetic algorithms (GA) for sketch-based software conceptual modeling. Firstly, a structure of object relation graphs (ORG) is constructed based on the definition of the spatial styles in UML. Secondly, a computational model is employed to describe the interrelation of sketchy UML objects and capture frequently occurred spatial styles possibly with fixed semantics through genetic algorithms. Experiments have proved the proposed method both effective and efficient.