Abstract: Focusing on the automatic recognition and extraction of plowland boundary in Jianghan Plain based on aerial images, the urban road network space research was introduced to make up for the defects of traditional farmland boundary extraction. The boundary of the plowland parcel in different forms was considered as roads in the urban road network space, and the road-like network space of plowland parcel boundary was constructed. The land parcel boundary was organized in a road-like network space environment, and an appropriate deep learning model was introduced to extract the global static and multi-morphological land parcel boundary. In general, based on the accumulation of previous achievements and the exploration of methods at this stage, a feasibility study had been made to solve the problem of automatic identification and extraction of cultivated land boundary in Jianghan Plain based on aerial images, which provided support and inspiration for other scholars to carry out research on farmland parcel boundary for plain agricultural applications.

Key words: plowland, farmland boundary extraction, road-like network, agricultural application