Indoor Scene Modeling in Power Distribution Rooms: A Focused Study Utilizing Attention CNN

The efficiency and accuracy of digital modeling in power distribution rooms are critical for infrastructure management. This paper introduces an innovative method for indoor scene modeling using Attention Convolutional Neural Networks (Attention CNN) to segment floor plans into functional components. The method addresses the limitations of traditional manual and CAD-based techniques, which are characterized by high costs, low efficiency, and inflexibility. The approach involves an Attention CNN that precisely identifies and segments key elements such as switchboards, distribution cabinets, and transformers from floor plans. Subsequent steps include image registration, coordinate calculation, model library adaptation, and 3D visualization. The method streamlines the modeling process and enhances the integration of model data with other information systems, offering significant advancements over existing technologies. The effectiveness of the method is demonstrated through experiments, showing substantial improvements in modeling speed, accuracy, and ease of updates.