Image based spare parts reconstruction for repairing vital infrastructure after disasters: Creating or ordering replica of spare parts for overhauling infrastructure

Abstract

From the very first hours after disasters such as earthquakes, hurricanes, floods, landslides, and tsunamis, survivors and rescue teams start repairing the devastated vital infrastructure. In doing so, spare parts are frequently needed, which cannot be delivered via destroyed roads or cannot be found in collapsed warehouses. In this work, we present an approach of how spare parts can be reconstructed out of images, and printed out by a normal or 3D printer. Therefore, a structure from motion algorithm is applied to a small number of images, showing the spare part from every direction, which can be captured by any kind of digital camera. Based upon the resulting dense 3D point clouds, a meshed 3D model is computed. After estimating the dimension of the real parts with a straightforward user interface, the meshed virtual model can be scaled correctly. In order to transfer the virtual 3D model back to the real world, our approach provides a 3D model format - for 3D printers - and a stacked paper format - for normal printers. In case no printers are available, the created 3D model can be transmitted via any communication network like GSM, digital radio, or internet to the next available printer or warehouse, nearby. For demonstrating our method, we experimentally reconstruct a gear wheel of a water pump. Finally, we discuss advantages, drawbacks and further steps-necessary for making our approach available. © 2016 IEEE.