Can someone give me an idea of how precisely I can measure the height of something (like a corn field) using a DJI X3? I need to be able to detect somewhat subtle surface differences based on variety, is this possible?
Not easily. Can to explain your objective a little more? Limitations in sfm / photogrammetry, particularly on corn, are going to have a big impact on what is possible, here. Corn has a big trade off in altitude. Low altitude doesn’t work well. Wind is also an issue - corn moves a lot…and falls over on the edges. Add to that sporadic cloud cover etc and you end up with a complex set of variables that it is very difficult to get a good, consistent and accurate result from…especially with an X3. Different camera systems (global shutter etc) offer better solutions, but radar / lidar is always going to yield the best results for this as it can see the ground through the canopy. If you can explain what you are looking to achieve I, or others on here, might have some ideas.
One tried and tested solution with sfm is with RTK gps correction and a pre-sowing baseline map. Although the ground moves a little and you have around 2cm positional accuracy, you can subtract baseline from key points in growth to give relatively accurate and consistent crop height and conveniently eradicate a lot of the ‘noise’. But you can’t (normally / practically) do it with a DJI Inspire and X3. If you need more accuracy than that then you need radar / lidar or an in-field sensor.
This is for plant genetics research. We have two major problems that require extensive time in the field: Stand counts and Plant height. From what I know about UAVs, the first is fairly easily accomplished with mapping, the second is the hard part.
The field is structured into a grid of short rows of different varieties with gaps between, so there are points to help assembly (not like a dense corn field). My ideal package would be something I could take images of the field when the corn is small for stand count, then take similar pictures at maturity to assess differences in plant height. These do not have to be down to the centimeter, but I would like to be able to rate each variety “High, medium, Low” or something similar. The stand count is my priority in this, but if it is possible to get the ability to assess differences in plant height also, I would really like that.
My plan is to use an X3 and possibly an X3 NDVI conversion to do some of my graduate research, but the ability to do stand count and plant height would be a huge savings for me. I will be paying for the UAV, data processing, etc all myself, so it needs to be a somewhat simple solution, and not ridiculously expensive.
Can anyone provide pointers on how to go about this?
Are you aware of the work at Rothamsted?
- Setup permanent GCPs (DroneDeploy support guys can help you when processing the GCPs).
- Take a base map after sowing
- Take an emergence map which you can do around 40m for good results on the X3 for stand count.
- Take pre-harvest / end trial map for plant height. Given your grids you should get good height data, but be aware of plot spacing and possible rope boundaries to prevent / limit crop fall-over and keep boundaries clear of encroachment because all of that will skew your results. Basically, where the standing corn ends you want a nice, clean edge and then nothing but soil on the plot boundary. Then, just map it on a still day and you should get pretty good results.
I’d recommend flying EACH map twice at two different altitudes. Say 40m and 80m. Earlier in the growth the lower altitude will give better results, latter the higher but by flying both you have the cross-reference should you need it.
Similarly with the x3 NDVI, although you’ll really struggle to get much useful out of that at 40m.
I’m doing research in agriculture for software I’m planning to build and I was wondering what decisions can be made based on stand counts, as well as plant heights, when growing corn. Any chance you can help me out?