Computer vision based steering path visualization of headland in a soybean field

Yuyang Ren¹Bo Zhang^2*Yang Li³Changhai Chen⁴Wenxiao Li²Yongcai Ma²

• ¹college of information and electrical engineering, Heilongjiang Bayi Agricultural University, Daqing, China
• ²Heilongjiang Bayi Agricultural University, Daqing, China
• ³Beijing Simulation Center, Beijing, China
• ⁴Heilongjiang Academy of Agricultural Sciences, Harbin, China

To address the insufficient accuracy of autonomous steering in soybean headland areas, this study proposes a dynamic navigation line visualization method based on deep learning and feature detection fusion, enhancing path planning capability for autopilot systems during the soybean V3-V8 stage. Firstly, the improved lightweight YOLO-PFL model was used for efficient headland detection (Precision was 94.100%, Recall was 92.700%, mAP@0.5 was 95.600%), with number of parameters and float point operations of 1.974 M and 4.816 GFLOPs, meeting embedded deployment requirements for agricultural machines. A 3D positioning model was built using binocular stereo vision, distance error was controlled within 2.000%, 4.000%, and 6.000% for ranges of 0.000-3.000 m, 3.000-7.000 m, and 7.000-10.000 m; Secondly, got interference-resistant crop row centerlines (the average error of orientation angle was -0.473°, with the negative sign indicating a tendency for the measured orientation to deviate slightly to the left of the reference axis, suggesting a small systematic bias, the Mean Absolute Error (MAE) was 3.309°) by first enhancing contours with HSV colour space conversion and morphological operations, then fitting feature points extracted from ROIs and crop row intersection area using the least squares method. This process solved centerline offset issues caused by straws, weeds, changes in illumination, and the presentation of holes or sticking areas; Finally, 3D positioning and orientation parameters were fused to generate circular arc paths in the world coordinate system, which were dynamically projected across the coordinate system to visualise the image plane navigation lines. Experiments show the method generates real-time steering paths with acceptable errors, offering a navigation reference for automatic wheeled machines in soybean fields and technical support the advancement of intelligent precision agriculture equipment.