Photos into Pointclouds ... Wizardry at its finest?  

Recap Photo (and the old free product 123D Catch) … one of (I think) the most exciting technologies available to try from Autodesk.

Photos into Pointclouds ... Wizardry at its finest?  

Many of you will have played with the Autodesk technology 123D Catch,  This was a free photogrammetry software that could turn your photos into a 3D visualisation.  Since those days Autodesk have developed the software into a new service called Recap Photo which can turn your photos into point clouds and meshes that you can use with other applications such as Inventor, Revit and Navisworks. This is an exciting idea, especially as the services or equipment used to laser scan a room or an object can be expensive and require some knowledge to get the best out of them.  On a small project it just isnt feasible and we need a smaller scale option - photogrammetry give us this option and Recap Photo makes it easy.

However, as engineers we are a skeptical bunch! Some questions I keep getting asked are “How accurate is Recap Photo?” and "It would never work in real life would it?"

Well a few years ago there was an article published by Loughborough University here in the UK. This article explains some of the testing they have done with Recap Photo, or as it started out life as 123D Catch, and how it compared to other reality capture methods.  It explained that during a survey of a cave system they managed to get an accuracy of 12mm use the software to convert the photos into a point cloud.  This however was based on the images they had taken.  More photos at a better resolution would improve the accuracy.  Not bad for some photos!

The software has moved on massively since those early days and now in principle, reconstruction is accurate within ±1 pixel of the input images. Accuracy and resolution are dependent on the ground sampling distance in the images and other image capture variables like the quality of the camera and lens system, camera shake, focus, etc. information on deviation on each input Ground Control Point.  But the accuracy can in theory meet that of laser scanning in many situations.

Late last year I watched a presentation from Autodesk and the US Bureau of Reclamation on their project at the Glen Canon Dam over in the states. As part of this project they combined data from many sources such as laser scans, photos, sonar data to generate one aggregated accurate model; there were as many as 900 scans all pulled into one accurate model!  The interesting thing I found was how much photogrammetry played a role in this project and how the various technologies were used for different situations.

If you really want to see how these different technologies can integrate then check out this blog post by Autodesk on their project: https://medium.com/autodesk-university/virtualizing-the-glen-canyon-dam-29c7d30e68c5

Recap Photo is available to all users of the Autodesk Design and Manufacturing Industry Collections.  I would really advice you take a look and see how it could fit into your sales > Engineering or project processes.

Blog

Machine Learning meets Carbon Analysis

17 April 2024

Factoring embodied emissions into building designs is a growing area of concern in the AEC sector, with many practices now investing considerable time and resource into low carbon design strategies.

Learn more
Blog

What and why is 2FA / MFA important?

15 April 2024

In 2023, over 1000 companies were hit worldwide by Ransomware and over 15 billion credentials were stolen and obtained globally through hacking and bating companies and users, causing havoc on an unfathomable scale. Read this blog to find out why Two Factor Authentication and Multi Factor Authentication is important.

Learn more