An efficient crop row detection method for agriculture robots

In this paper an efficient crop row detection method is proposed for vision-based navigation for agriculture robots. In the proposed method, no low level features (such as edges and middle lines of the images) are needed. So the complex algorithms for edging and matching (e.g. the Hough transform) are avoided, which greatly saves the computation loads. Instead, a flexible quadrangle is defined to detect the crop rows. The proposed method moves, extends or shrinks the flexible quadrangle to localise the crop rows in the captured frames. The experiments demonstrate that the proposed method is effective with high time efficiency and detection accuracy.