Research progress in farmland boundary extraction methods based on remote sensing image
The automated and accurate extraction of farmland boundaries plays a crucial role in effectively quantifying agricultural land resources and setting relevant regulatory policies.It is essential to the development of modern agriculture and smart agriculture.This paper compiles commonly used remote sensing data across various extraction scales and outlining the development of boundary extraction techniques from unsupervised to supervised methods.The edge-based,region-based,hybrid and deep learning methods were comprehensively summarized and compared.Deep learning,a supervised extraction technique,is closely linked to advancements in remote sensing technology.Accuracy assessment involves reference image acquisition and evaluation methods.Current challenges include the underutilization and limited exploration of farmland boundary features,the lack of publicly available and real-time datasets,the scarcity of research focused on smallholder areas,and the limited applicability of existing extraction methods.In order to overcome these challenges,the future directions of research were proposed,such as the integrated utilization of different farmland boundary features,the establishment of data-sharing platforms for comprehensive farmland boundary data,the application of cloud computing in farmland boundary analysis,the development of specialized extraction methods for smallholder areas,and the integration of spatiotemporal data from multiple sources.