A U-shaped Corridor-based Single Model Approach for Furrow Detection Throughout Agricultural Work Cycle

The automation of open-field agriculture is highly important in addressing the decline in rural populations and food security. In open-field agriculture, to prevent crop damage, the system recognizes the ridges or furrow where crops are planted and performs tasks while navigating. However, previous camera or LiDAR-based systems face the issue of needing to create or retrain models for different ridge environments altered by agricultural work. This study proposes a method that defines the morphological features of furrows and uses a single trained model, allowing it to be applied regardless of the ridge conditions altered by processes. Additionally, the method involves connecting the central points of each recognition model to create a navigation path for the robot based on furrow detection. To validate the proposed method, it was applied to two ridges with different depths under dry, wet, mulching, and sowing conditions. Consequently, the average recognition rates for the two ridges with different depths were 97.36% and 87.13%. This means that it can be applied in a variety of field farming environments.