Hi all,

I'm geoscientist and drone pilot, but have recently started to improve my GIS and remote sensing skills by attending some advanced GIS courses. One of the courses requires me to complete a freely chosen capstone project to show my skills, and that's why I'm posting this:

As I'm really interested in UAVs, multispectral data, and automated classification (or feature detection) I thought automated plant detection and classification from UAV imagery might be a cool project to work on. My idea is to use multispectral drone imagery (e.g. taken with a MicaSense RedEdge or similar camera) of one or multiple plantations and try to develop a workflow in ArcGIS that automatically identifies individual trees/bush/plants (whatever is growing on the plantation), counts them, and classifies them by their health. This is certainly nothing new but a nice project to test my skills.

Problem is, I neither have a plantation nor a suitable multispectral camera.

Would somebody here be willing to provide me with some UAV multispectral imagery from plantations that I can use for my project? At the moment I don't really care where the plantation is located or what kind of plants are growing there. However, it should be something with a little bit of spacing between individual plants (hence no crop fields), and at best some of the plants are more healthy than others.

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • Hi Philipp,

    Sorry for taking to long to get back to you.

    Please, send me a private message such I can pass the link to the raw data set (spectral + rgb).

    Regards,

    Wilson

    3702371256?profile=original

  • Hi guys,

    sorry for the late reply.

    Someone gave me some data that I used to test my skills. Results are certainly not perfect and have some false detections, especially where trees are very closely spaced and form some sort of big hedge. A GIS expert might get more out of it.

    Below are two images, the first shows the delineated tree crowns and a height classification, the second has the NDVI.

    If someone has comparable data I'd be happy to test my model with it. I need a DSM to extract the trees, and spectral data to assign plant health. :-)

    3702371042?profile=original

    3702371121?profile=original

  • Moderator

    Hi Philipp, please send me an personal message, i will provide some true NDVI samples for you

  • I'll send link to download soon.

    Philipp said:

    Hi Wilson,

    the dataset looks really nice, exactly what I was thinking of, but I would need the raw data, i.e. the R G B and NIR bands.

  • Hi Neil,

    that's a nice shoot, but can you identify individual trees on the images? An individual tree should at least cover a couple of pixels, otherwise the identification will be really hard...

  • Hi Phillip

    I have several sets of data available (using an Event 38 modified camera co-located with a normal unmodified camera) and I plan on doing some more testing in the near future. If you have a specific need for a certain type of subject send me a pm and hopefully we can work together. I use a full size aircraft for this so there is no need to arrange access or permission from the land owner for small data sets.

    Attached is a "raw" image from the event 38 camera and screen shots of the associated "processed" image, and the normal camera image is too large for upload (51 Megapixel camera). 

    3702368514?profile=original

    3702368497?profile=original

    Kind regards

    Neil

    IMG_7267.JPG

  • Thanks Will, I'll send you a message! :-)

  • Hi Wilson,

    the dataset looks really nice, exactly what I was thinking of, but I would need the raw data, i.e. the R G B and NIR bands.

  • Hi Philipp, I could provide you with imagery of a citrus grove, varying health, and several plant species, e.g. pine trees.  Send me a message if you are interested.

    Will

  • Hi Philipp,

    Are you willing to try in a data link this one (NDVI sample, after classification)?

    3702368646?profile=original

This reply was deleted.