Automatic image analysis and generation of digital residential floor plans

点击翻译按钮获取中文摘要

The design of electrical plans plays an important role in all residential and commercial construction/renovation projects. Electrical plans are created manually by electrical designers who must follow the Canadian Electrical Code and industry best practices. The designers decide what type of electrical devices (e.g. light switches, outlets, etc.) are required and where they should be located based on the architectural floor plans. This is a time consuming task, which is prone to errors and inconsistencies. There is no intelligent, context-sensitive software tool able to help designers expedite and streamline the electrical plan design process.In this research, we propose to automate the electrical design process through computer vision methods. We will develop an automated system that will allow a computer to analyze and understand images of architectural floor plans and their embedded data by processing visual information. Relevant data, such as architectural symbols, building structures, and room types and shapes, will be retrieved automatically. This information will be used to generate the electrical plans by determining the required electrical devices as well as their optimal placement. The system will be able to learn and adapt to a considerable number of graphical notation types and conventions, as there is no standardization in the floor plans.Integrating the proposed system into the current electrical design workflow will bring many advantages to the Canadian industry. It will ensure consistency in the designs, increase productivity, generate cost-effective solutions, and bring the latest computer technologies to the industry for a clear competitive edge. Moreover, the automatic analysis of architectural floor plans paves the way for many other applications besides electrical design, such as plumbing design. The novelty of this project will also allow to fill many existing gaps in the research on graphics-rich document analysis.