Automated robust crop-row detection in maize fields based on position clustering algorithm and shortest path method

Crop row detection is critical for precision agriculture and automatic navigation. In this paper, a novel automatic and robust crop row detection method is proposed for maize fields based on images acquired from a vision system. As the image quality is easily affected by weed pressure and gaps in the crop rows, the proposed method was designed with the required robustness in order to deal with these undesirable conditions, and it consists of three sequentially linked phases: image segmentation, feature points extraction, and crop row detection. The image segmentation is based on the application of a modified vegetation index and double thresholding combining the Otsu method with the particle swarm optimization, thus achieving a separation between the weeds and crops. During the procedure of crop row detection, the position clustering algorithm and shortest path method were applied successively to confirm the final clustered feature point set. Finally, a linear regression method based on least squares was employed to fit the crop rows. The experimental results show that the detection accuracy of this proposed method is 0.5°, which out-performs the classical approach based on the Hough transform.