Vision-based field inspection of concrete reinforcing bars

Concrete reinforcing bars should be accurately placed in the positions shown on the construction drawings, adequately tied and supported before concrete is placed. These elements should be further secured against displacements within the tolerances recommended by project specifications. Ensuring compliance with contract documents and the building code applicable to the project under construction requires photographic documentation and close visual examination by field inspectors. Although inspection procedures are repetitive for every jobsite, the manual inspection methods are time-consuming and non-systematic. Moreover, the current practice of field inspectors walking into rebar cages and footings for close assessments can be a potential safety hazard on jobsites and can damage the integrity of the structure. To minimize the challenges of the current practice, this paper proposes a computer vision-based method for field inspection. In the proposed method, a field inspector can carefully walk around a rebar cage and take a complete collection of images from the underlying structure. Using a vision-based 3D reconstruction pipeline of Structure-from-Motion and Multi-view Stereo algorithms, a dense 3D point cloud model will be generated. Using an algorithm that maps and generates a density histogram of points, the locations and configuration of the rebars are identified. Finally, the spacings between rebars are calculated for field inspection. Experimental results on data collection, analysis, and visualization components of the proposed rebar inspection method is presented. These results show the promise of applying this low-cost approach in practice.