CODEWITHSUNDEEP
pythonopencvconv-neural-network
fferrin
fferrin

Reputation: 908

Detecting farm crop boundaries from satellite images

I want to detect crop areas from satellite images.

If we copy any image of google maps that contains a farm land, you can easily see the divisions, but using OpenCV is not so easy for me.

First, I convert the image from RGB to gray and apply a Gaussian filter.

enter image description here

When I apply a threshold to the image, I get many irregularities that make it difficult to separate the different areas.

enter image description here

I had thought about clicking somewhere and use some flood algorithm to detect at least the edges of the terrain clicked. Another (more advanced) option is to use CNN but I still do not have much experience in that and I would like to start with the basics.

Any suggestions?

Upvotes: 1

Views: 2932

Answers (1)

Ian A McElhenny
Ian A McElhenny

Reputation: 1027

Consider,
1) Breaking the image up into many small images
2) Increasing the strength of your gaussian filter to remove noise
3) Using hough line detector to get edges from the binary image (black and white)

You will need to get the image as clean as possible. By clean I mean, make sure the edges are very clean with minimal noise on the inside of the fields.

Lastly I would suggest using an edge detector with a hough line detector. Look into canny edge detection.

Upvotes: 1

Related Questions