TY  - JOUR
T1  - Farmland Segmentation for Autonomous Agricultural Machinery
AU  - Bae, Na Yeon 
AU  - Choi, Sung Kyun 
AU  - Han, Dong Seog 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.4.587
KW  - Segmentation
KW  - D-block
KW  - DG-block
KW  - pixel shuffle
KW  - semantic segmentation
AB  - Smart agriculture leverages information and communication technology in farming to enable automation, providing a sustainable solution to challenges such as climate change and an aging population. Recently, there has been active research on agricultural automation by integrating autonomous driving technology into key agricultural equipment, such as tractors and rice planters. This paper proposes a deep learning architecture to distinguish cultivable land. Using images of farmland captured by drones, we construct a dataset and aim to classify areas such as fields, edges, and roads with a lightweight deep learning model. This paper proposes a deep learning model that refines image regions using a DG-block (Dilated Group Convolution-block) and pixel shuffle. The proposed system demonstrates performance with an mIOU of 78.4%, an accuracy of 77.7%, and an inference time of 50ms.