Regular visitors to the News section may remember that Simon Brown (Deep3D) and Marcus Blatchford had a very informal project running to map as much of the National Diving & Activity Centre as they could.
If you want to read more about how the model was created, do please read on. But if the model is all you seek, here it is:-
Each of the attractions are annotated, and some have clickable links to take you into the details.
A Big Project
For those who have never visited this inland dive site, its worth pointing out that the dive site is massive and has a lot of widely dispersed diver attractions. Linking up attractions such as the Land Rover, Stalwart, Trawler and the Wessex Helicopter was progressing, albeit slowly. Part of the site was already joined, merging data from several dives:
But there were some huge gaps, mostly between the Land Rover and Stalwart, and from the Abbott Self Propelled Gun to the Trawler. And beyond the Wessex helicopter was unrecorded territory.
Marcus decided to devote a dive to filling the gaps, not capturing the detail of the attractions but shooting just enough of them to allow them to be inserted into the overall model.
The volume of images now needed to process the basis of the complete model was starting to hit the hardware limits. The model has close to 9000 images to manage and process. Working the images in a single model is a quick way to find the hardware limits.
Thankfully our friends at Agisoft had already authored a handy python script for use in Photoscan Pro that can dice up a complex model into neat chunks. Each section can then be processed as a single model before being merged back into one. Here’s a screen shot after the point cloud (330 million points) had been split into four:
Using this custom tool allowed the model to run in separate batches. This meant we could build more – much more – of the site in a single model. This technique will serve The Shipwreck Project well in the coming year.
A big thanks to Agisoft and Marcus for providing the tools and data to build and fill the gaps. Here’s the result:-
Now we have the model, the next step is to create an orthophoto of the site. This may take a while…