Vision based guidance line extraction for autonomous weed control robot in paddy field

Abstract

Weed control in rice farming is important for increasing yield. A guidance line can enable a robot to follow the crop row without damaging crops using a successful weed control-equipped vision camera. The guidance line is generally extracted from an image of the rice rows, and the accuracy of the guidance line is affected by the morphological characteristics of the crop. In the case of rice plants, extracting the guidance line precisely is difficult because rice leaves are laid out in multiple directions. This paper proposes a new guidance line extraction algorithm to improve the accuracy of autonomous weeding robots in paddy fields. The proposed algorithm determines the central region of the rice row using the morphological characteristics of those leaves that normally converge with the direction of the central stem region using a simple image processing operation. The guidance line is extracted from the intersection points of virtual straight lines using a modified robust regression. The virtual straight line is represented as the extended line from each segmented straight line that was created on the edges of the rice plants in the image using the Hough transform. The proposed algorithm was observed to have good performance experimentally, with a high accuracy of less than 1°.