Stereovision-based ridge-furrow detection and tracking for auto-guided cultivator

This paper proposes a stereovision-based auto-guidance method for a riding-type cultivator to track the inter-rows between ridges and furrows. A stereo camera perceives the three-dimensional (3D) structure of field surfaces and enables autonomous agricultural vehicles to determine navigation paths. However, when developing an efficient ridge-furrow classification and guidance-line-extraction algorithm for outdoor applications, the stereo camera must overcome the irregular vehicle movements caused by uneven surfaces and errors caused by strong sunlight. When moving on uneven surfaces, pitch and roll oscillations of the stereo camera would cause disparity variations corresponding to the ground plane. Furthermore, due to the image artifacts caused by strong outdoor sunlight, the boundary discontinuity between the objects would disturb the robust detection of the guidance line. The primary contributions of this study were the development of a compensation method that responds to the dynamic pitch and roll movements and a robust guidance-line extraction method that responds to image artifacts. The developed algorithm could classify ridges and furrows with an accuracy above 90% under outdoor conditions. On-site testing with auto-guided riding-type cultivators using a stereo camera confirmed that the developed ridge-tracking algorithm is useful for real-world agricultural applications and revealed a lateral deviation from the average root mean square error (RMSE) of 2.5 cm in a flat field and 6.2 cm in a hillside field.