Automatic Acquisition of Rules for 3D Reconstruction of Construction Components

Aiming at the rule-based recognition system for architectural structure drawings, this paper presents a sample-based method that will automatically extract the geometric features of architectural objects and convert the features into recognition rules. When users designate the first sample of one type of objects, this method automatically analyzes and extracts the features useful for drawing recognition. Aided by a little interactive operation, recognition rules for this type of objects are generated. When a new sample of one object is designated through automatically comparing the new features and the old, this method analyzes and modifies the features of this object, then generates new rules. Along with new object types or new representations of old types are processed, rules of the recognition system are perfected continuously without modifying the program. This method can bring to the drawing recognition system higher self-adaptability and practicability.