Image analysis of plant growth

martindemel

Has anyone come across companies (guess mainly agrecultural ones) who are using camaeras to make pictures of their plants and then analyse the number and area of leaves to detect the final weight of a plant? I'm sure this is done in some way, but do you know of anyone using JMP for this? 

 

Compared to a image for defect analysis/3D foot prints/surface inhomogenities there is the issue of overlapping or hidden leaves. So the standard approach probably won't work.I guess one needs more than one camera and some other information about the plant itself (symmetry, number of levels over time/height, average number of branches, ...). Any experiences or contacts I could share with my customer?

 

Thanks,

Martin

 

@JohnPonte 

JohnPonte

Hi Martin,

 

Once I was asked if JMP could be use to analyze images of forest area. A drone was flown over an area filled with trees. In that case, if I could identify the ground (through the trees) then I could estimate the area of foliage vs. the ground. I did this similarly to how I handled the elephant foot with sores - Image analysis of an elephant's foot in JMP . But I don't think thats what you are asking for. In that case I was calculating the area covered by the trees collectively. I was not counting individual leaves, or even trees. For your case you would have to identify individual leaves, even leaves that are obscured by other leaves. That becomes a much harder problem. In the forest area case I only had an aerial view which is two-dimensional. For your case, you would need a three-dimensional representation of a plant to be able to count the leaves underneath the top layer of leaves.

 

It seems like the goal is to estimate the weight of the plant. I wonder if you could do this without actually counting all the leaves but rather just calculating the foliage coverage and perhaps the height of the plant. The foliage coverage would represent how "full" the plant is and the height would be an assumption about how many layers of underlying leaves there are. The taller the plant, the more leaves there are (assumption). In the end, it is always going to be an estimate, no matter what approach you take.

 

I think the problem, as you describe it, would be fairly difficult in JMP. I wonder if there is a way to redefine the solution such that it could be more easily done in JMP. If you have some images to share I would be happy to take a look at them.

 

Good Luck!

John Ponte