CT scanning has become an important way of seeing how plant roots grow in soils. If we can see the structure of roots and surrounding soil without destroying the system we are looking at, we can learn a lot about processes which might help us to improve agriculture.
Unfortunately, getting information out of the images is difficult. This is partly because each of the 3D images from the CT scanner can be 30 GB in size. It is also partly because it isn’t easy to train a computer to spot automatically what we can see easily with human vision. For example, teaching a computer to differentiate between different parts of the soil – and to do this reliably – is not straightforward.
In this paper, we looked at making it easier to get 3D shapes out of data from CT scanners. We put samples of different soils into a CT scanner to get 3D images, and then wrote a program which automatically extracts the 3D pore network.
This kind of work is important in developing the tools we need to make 3D imaging methods (like CT scanning) effective on a large scale to help improve crop breeding and agriculture.
