3D Modelling using OpenDroneMap

We made use of OpenDroneMap to map out an area. It is a tool for postprocessing aerial imagery, producing geographic data including orthophotos, point clouds and textured meshes. It makes use of OpenCV to carry out very complex and processor-intensive calculations, through the execution of various algorithms such as DoG, Harris corner detection and bounding volume hierarchy calculations. By taking aerial imagery, or any imagery with regards to a common 3D object.

This algorithm goes far beyond the algorithm developed for image stitching, as the image stitching done in the previous section was only for 2D image stitching, and only ever two images at a time. However, this is a far more well-refined and sophisticated algorithm as it is able to extrapolate 3D data points from 2D images, and also without stereo cameras.

Once the script completes, it produces multiple folders with many files, some of which are the meshes and textured meshes that we would need to show the 3D model. In the above figure it can be seen that it generated a point cloud, a non-textured mesh, and a textured mesh. These meshes are very useful to observe the surrounding area the drone flew in. In the case of search and rescue, this is a vital feature to have as when an area is completely destroyed, it is vital to be able to obtain a quick model of the area as any existing maps or physical representations of the area are likely to be inaccurate and outdated, in terms of bounding volume hierarchy representations.

More about OpenDroneMap can be read here: http://opendronemap.github.io/odm/pages/about.html