Computer vision holds enormous promise for the construction sector. It can examine numerous data types such as point clouds, pictures, and video recordings from construction sites in real-time, thanks to its object recognition capability. In terms of safety, it can identify structural deterioration through segmentation techniques and monitor security camera video to identify worker and machinery movements at construction sites while assuring worker safety and optimizing worksite activities. Leveraging BIM data over the whole building life-cycle, we aim to improve processes in the maintenance and operation phase by generating so-called digital twins. Computer vision methods may be used to create as-built models, which can be enriched into digital twins and be compared against requirements. Our research is split into two primary categories: machine learning approaches in digital twin generation and computer vision techniques in the safety monitoring of infrastructures.
Artificial Intelligence techniques and methods may be utilized in structural health monitoring for buildings and infrastructure for various applications. In our group, research projects include developing automated visual inspection in concrete structures and employing AI and technologies to assure safety in construction sites.
In this research field, we deal with the automated generation of BIM models for existing buildings and infrastructure, so-called as-built BIM models, using AI methods. We use raw data such as point clouds, images, 2D plans, and construction documentation as training data.