How can I detect a row of crops (agricultural plants) with lots of foliage?

Question

I need to detect a row of crops and extract lines from a live feed for guiding an agricultural vehicle in the space between the rows. The problem is that the plants are very tall and the leaves get in the way as well like so:

Tall plants in a row

So far I've tried Blurring with large apertures + Canny + Thresholding + Probabilistic Hough Transform and played around with the parameter values for each of these. Nothing seems to work.

Also, the grain heads on top of the plant are in a different shade of green than the rest of the plant. But I can't figure out how to filter those out.

All of the existing literature involving crop row detection involves small plants like soybean, sugarbeet.

Does anyone have any advice? Would it be helpful if the images were taken from further above?

what about color detection (hsv) and perspective correction? — Micka, Jun 16 '17 at 20:37
Will the robot be taller than the plants, and so have a viewpoint similar to what you show here? Or will it run on the ground down lower in between the plants? Point of view matters if you want to use photo imagery to guide motion. — TomServo, Jul 10 '17 at 21:43

score 0 · Answer 1 · answered Jul 01 '17 at 19:20

Yes, try taking the images from further above. Start by preprocessing your image. This includes de-noising and segmenting the image. After the image is segmented, calculate the centroid coordinates of each region. Optionally, you can then calculate the centerline of each row using, say, a Randomized Hough Transform. Take a look at this paper.

If you want something a little simpler, try using an HSV algorithm. Hope this helps.

How can I detect a row of crops (agricultural plants) with lots of foliage?

1 Answers1