Automatic floor plan detection and recognition

We present a system that detects floor plan automatically and realistically populated by a variety of objects of walls and windows. Given examples of floor plan, our system extracts, in advance, bearing wall, setting others objects which are not bearing wall into a non-bearing walls set. And then, to find contours in the non-bearing walls set. It recognize windows from these contours. The left objects of the set will to be identified walls including with the original bearing walls. The last step is to disintegrate wall into independent rectangular one by one. We demonstrate that our system can handle multiple realistic floor plan and, through decomposing and rebuilding, recognize walls, windows of a floor plan image. Based on high resolution images downloaded from Baidu, the experimental result shows that the average recognition rate of the proposed method is 90.21%, which proves the effectiveness of the proposed method.