NutLoose
28th Nov 2022, 10:22
I was going to post in the Ukraine thread but it deserves it's own thread.
They 3D scan the items in use by the military then the software they have developed will detect munitions on the ground from those scans.
https://vframe.io/about/collaborations/t4t
Various trials under different conditions
https://vframe.io/about/collaborations/t4t
unmanned_droid
28th Nov 2022, 23:35
The idea here is to be able to take photographs or videos of an area and have a suitably trained algorithm scan the images looking for pattern/shape/colour/tone detection based on its library of known images. The library can have 1 or many entries. The library is required as it acts as 'training' for the algorithm. This project creates the training data for the algorithm.
Someone like myself could create a drone configuration that is able to record very high resolution images and take accurate and geotagged photos of a site, and then have the algorithm scan the photos and plot points or areas it thinks contains objects within the library. Clearly this works best for fully exposed objects, and there is some amount of partial matching available to, say, find a munition half in the ground.
Key factors here are that no algorithm can be considered 100% successful, submerged munitions are not likely to be found unless they are being looked for (ground sign for example) and it requires a thorough coverage ideally with an accurate positioning system (RTK GPS for example).
Potentially, this could also be done from vehicles and on foot.
PFM-1s have also been used fairly extensively, so hopefully that's what's up next for this group.
Good work from them!