Automatic indoor construction process monitoring for tiles based on BIM and computer vision

For project managers, accurate, timely, and intuitive information is the key to successful decision-making during progress control at a construction site. This paper presents a method that combines computer vision with building information modeling (BIM) for automated progress monitoring of tiles. This method can automatically and accurately measure the in-built progress information of a construction site and transmit real-time progress information to the cloud in a visualized form. With the in-built and real-time progress information, project managers can know the progress of construction site in time and make decisions easily. The proposed method includes several modules. First, an image database is built with thousands of tile images, and a variety of local binary patterns (LBPs) feature extraction methods, and support vector machines (SVMs) are used to train a tile classifier with satisfactory performance, with an accuracy of 91.17%. The purpose of the first module is to construct a mathematical feature of an image and to train a classification algorithm according to this feature, so that tiles in images can be identified. The improved edge detection algorithm detects the boundaries of completed tiles in given images. Afterward, the boundary line coordinates are converted from image pixel coordinates to a real-world coordinate system through camera calibration. Next, using the information from the camera location and room profile information extracted from the BIM model, the actual tile area can be calculated automatically. Finally, the in-built progress is highlighted in the room floor plan, and the result is delivered to the BIM cloud simultaneously. The proposed method was tested at a real indoor construction site. The experimental results indicate that the method can effectively carry out real-time automatic quantity calculations.