Support Website Contact Support Blog

Greenhouse gantry system mosaic


Our team is trying to image plant beds in a greenhouse to monitor plant growth and health in contained environment. To this end we have built a gantry system from which we hang our RGB and NIR cameras. As the drone flies, the gantry rolls. Overlap is good. However, artefacts are frequent in the final orthomosaic  because all the images have the same geolocation and we have no GCPs. The plant bed measures roughly 2 metres wide by 6.45 metres long. I am wondering if we can create a custom spatial reference so that we can use GCPs and eliminate the artefacts. It is important that perspective is corrected; therefore, photo stitching is not a solution for this project. I thought about the possibilities of selecting arbitrary projection for both the images and the GCPs or using GCPs as tie points but not sure that would help at all. Please advise.

Hi Gary,

You can work in an arbitrary coordinate system and use GCPs.

Another possibility would be to add scale constraints to the project which would be probably faster and I think it will be equivalent for this purpose too. You can find information on how to do it here.

However, I do not think the artifacts are generated due to a wrong geolocation. If Step 1 produces a good result and the block is well connected, the mosaic should be fine.

Please open a ticket in support if you want us to analyze your project in more detail.

Thank you very much.

Hi Daniel, I am actually the same person only under a different account now, as Gary no longer works in our group. Thank you for your answer - I ended up using scale constraints and it did the job. However, the resampling from this procedure also succeeded in lowering the quality of the resulting mosaic - I wonder if there is a workaround for this?

In addition, I would also like to process my micasense images of the same greenhouse bench but am not permitted to select the Ag Multispectral template as once again I do not have geolocation on my images. I need to produce a reflectance map so that we can analyze the plants using indices etc. with the infrared bands. Any solutions? Thank you.

Hello Anne,

The scale constraint should not end with a lower quality, it is just a new computation of coordinates, any resampling should be applied.

Anyway, why did not you use the GCPs?

As for the micasense camera, you can select the settings that you wish, there is no need to select a template. Just select what you want.

Anyway, if you do not have geotags, GCPs will be of great help in this case. Try to measure them in beforehand.

Let me know if this helps.