Paddy Rows Detection in "Legowo" Paddy Field Based on YOLO and Hough Transform

Paddy is a crucial commodity in Indonesia, where rice serves as a primary food source for its population. The "Jajar Legowo" planting system has emerged as a promising innovation in rice farming. However, activities such as weeding and planting are still primarily performed manually, which is inefficient and could be optimized with autonomous systems like smart tractors. For such systems, reliable movement along paddy rows is essential. This paper presents a method for paddy row detection using a combination of YOLOv4 and the Hough Transform. YOLOv4 is employed to detect the coordinates of paddy positions, which are then processed by the Hough Transform to generate virtual lines representing the paddy rows. The proposed model achieves a detection speed of 21 fps and a mean absolute error of 2.2 in row direction estimation. These results demonstrate the potential for automating paddy row detection and improving efficiency in agricultural practices.