Post-disaster robotic building assessment: Automated 3D crack detection from image-based reconstructions

Natural disasters can compromise the structural integrity of buildings. Three dimensional post-disaster building models, created using image-based 3D reconstruction techniques, would help early responders assess the extent of the damage. Image-based 3D reconstructions are advantageous as they only require a camera and tens of pictures to create a 3D mesh model. Under some circumstances responders can collect these images; however, if a building is suspected to have sustained significant damage, it may be neither safe nor straightforward to examine its structural integrity. Therefore, an approach is needed which allows responders to remain away from compromised buildings while gathering images for a 3D model. We propose the use of a small ground robot to take pictures of building elements remotely. Once the images are collected, an element model is created using a pipeline of Structure from Motion, dense reconstruction, and surface modeling. This model is compared to a pre-existing CAD model or examined to assess the likelihood of any structural damage. We also demonstrate a new automatic 3D crack-detection algorithm to assist examiners in identifying structural defects. Our preliminary results show that this process can successfully identify numerous cracks in the 3D reconstruction of a sample concrete column. The perceived benefits of the proposed method in a post-disaster situation are also discussed in detail.