This article describes how to create a Model Target by capturing a physical object using photogrammetry. This does not require dedicated hardware but present its own limitations. Please refer to this guide for an overview of available options to create Model Targets by scanning physical objects.
Firstly, you should select a software that is suitable for your needs. There are a variety available that each specializes in different fields of photogrammetry scans. In this article, we have selected two that accomplishes the mesh generation and texturing to a satisfactory level.
- Meshroom is an open source 3D Reconstruction Software based on the AliceVision framework. It is currently available for Windows, Mac and Linux operating systems. Follow their installation guide here.
NOTE: Mac users need to take additional installation steps. Please consult their guide for more information.
- RealityCapture by CapturingReality is a professional photogrammetry software with a high level and fast generation of models but it is available only for Windows. See their pricing lists here. The installation guide will appear in their application download menu and when you first launch the application.
NOTE: An activation step to verify your license is necessary when you first launch the software.
Please consult each software’s hardware requirements:
The following sections are directed on procuring a proper model for Model Target generation and they will not present all the necessary steps for using Meshroom or RealityCapture. Therefore, we advise you to familiarize yourself with the interface and workflow of either of the two software before going further.
Meshroom’s beginners guide is available here.
RealityCapture’s beginners guide is available here.
Image Capturing Guide
We recommend that you consult Meshroom’s and RealityCapture’s own guides on capturing images that are suitable for their software. You can find their guides at Meshroom’s Capturing Guide and RealityCapture’s Taking Pictures Guide.
However, depending on the size of the object, taking at least 100+ images of the object is recommended; covering all sides and at different angles and heights. Additionally, if the object is suitable, take images of both its entirety and of up-close details. It is better to have too many images than too few!
NOTE: Generally speaking, photogrammetry uses the background on the photographs as an aid to stitch and align each image together in a 3D reconstruction environment. Therefore, do not rotate or move the object while photographing it and make sure the background has enough detail.
NOTE: Alternatively, a seamless background and a rotating disk would allow you to reposition the object and in that way capture the object from all its sides, but the software will likely have more difficulty in positioning the images in its 3D reconstruction environment. This approach is most suitable for small objects. Please see the Meshroom topic on Turntables and RealityCapture’s discussion for more information.
Photogrammetry General Workflow
Meshroom and RealityCapture have different User Interfaces and tools when it comes to creating models from images. For example, Meshroom uses a node system for users that allows them to edit and iterate single stages in the process, whereas RealityCapture provides tabs and menus for procedural reconstruction stages.
Nonetheless, both follow a similar general workflow which we present here:
- Import group of images into the software and save the project.
- Create point cloud from images (Meshroom: StructureFromMotion node, RealityCapture: Align).
- It may be the case that some editing is required here. I.e. add additional images of a part of the object that was not captured well, or the ground plane needs to be defined.
- Create mesh from the point cloud (Meshroom: Meshing node, RealityCapture: Normal Detail).
- Texture the model (Meshroom: Texturing node, RealityCapture: Texture)
- Export Model for simplification and clean-up.
A model generated through the photogrammetric method requires additional editing. Firstly, depending on the final set up of your reconstruction process, the model might include a surrounding mesh that has to be removed. The model might also be in such high detail of both vertex count and textures that it is over 100MB in size; both should be reduced to avoid errors during the Model Target generation process in the Model Target Generator (MTG). And lastly, the model should be re-scaled to match that of the physical object.
Use your preferred mesh editor software to prepare the model for the MTG. Please refer to the Model Target Generator User Guide for information on supported file formats and general model preparation.
NOTE: Some aspects of model simplification can also be performed in the respective photogrammetry software. Please consult their documentation and practices on doing so. Meshroom Model Simplification and RealityCapture Simplify tool.
Model from the Meshroom reconstruction process before and after model clean-up.
When the model is ready, move on to the Create a Model Target section in How to Create a Model Target from a 3D scan.