A Method for Automatically Generating CAD Shelf Maps Based on Image Processing

Generally, the floorplan of a super store is manually drawn based on the multi-layered CAD shelf maps. This approach is accordingly ineffective, time-consumed, and easily subject to error. This paper presents a method for automatically identifying and localizing objects on CAD shelf maps by imaging processing. To obtain the map characteristics, the CAD file can be initially converted into a pixel based binary image. Features such as cross points, rectangular shape can be identified respectively by line-detection techniques and double-calculation methods with a certain threshold. Through this way, the pixel coordinates, lengths, and widths of all the objects in the image can be determined. The pixel dimensions of the objects can be converted to real-world dimensions by multiplying them with a scaling factor. This method is essential for indoor localization work and can be widely used for different large indoor scenarios such as stores or warehouses.