Robotic crop row tracking around weeds using cereal-specific features

Crop row following is especially challenging in narrow cereal crops, such as wheat. Separation between plants within a disappears at an early growth stage, and canopy closure rows, when leaves from different rows start to occlude each other, occurs three four months after the crop emerges. Canopy makes it identify separate through computer vision clear lanes become obscured. Cereal crops are grass species so their have predictable shape orientation. We introduce image processing pipeline which exploits track rows. The key observation exploited that leaf orientations tend be vertical along horizontal due location of stems Adaptive mean-shift clustering on Hough line segments then used obtain lane centroids, followed by nearest neighbor data association creating candidates 2D space. Lane parameters fit with linear regression Kalman filter for tracking frames. method achieves sub-50 mm accuracy sufficient placing typical agri-robot’s wheels real-world, early-growth wheat drive them, long seeded wider spacing 180 80 wheel width robot.