The 14th Asia-Pacific Agricultural Information Technology Association International Conference (APFITA2024) was held in Tsukuba, Japan from Nov 6 to 8. Focusing on the theme "Frontiers of Food Systems Originating from Sustainable Agricultural Production in a Data-Driven Society," the conference served as a pivotal platform for global scholars from 20 countries and regions  like China, Japan, Australia, South Korea, Thailand, and Italy to exchange cutting-edge technologies and foster innovation cooperation. The aim was to shape the current and future landscape of agricultural and food production system through collaborative efforts.

Dr. Duan Yulin, Deputy Researcher of the Innovation Team of Smart Agriculture at Institute of Agricultural Resources and Regional Planning (IARRP) of the Chinese Academy of Agricultural Sciences (CAAS), and Professor Seishi Ninomiya of the University of Tokyo, Japan were invited by the organizing committee of APFITA2024 to present a report titled "A Sample Free Framework for Precise Crop Type Mapping," showcasing the latest research achievements in remote sensing crop classification at ARRP.

Obtaining high-quality ground truth sample data is essential  for accurate crop classification and remote sensing mapping. However, the current method of sample acquisition often relies on extensive ground surveys and visual interpretation,  this poses challenges  for large-scale applications due to the time and effort involved. In recent years, many scholars have been dedicated to researching automated sample acquisition technologies. However, the core challenge in this field remains how to generate high-quality samples with spatio-temporal and spectral representativeness for various crop remote sensing classification tasks using existing prior knowledge. In the presentation, a sample self-inference generation method was proposed based on sample stratification strategies, considering sample spatial balance and crop spatio-temporal change characteristics, and gradually refining sample quality from coarse to fine selection. This method enhances the reliability and quality of the sample dataset, enabling precise crop classification extraction.

Experiments conducted across the three northeastern provinces of China demonstrated that, without using any ground samples or visual interpretation sample data, the crop mapping statistical results of this method were strongly correlated with data from the National Bureau of Statistics. The average correlation coefficients were 0.98 for rice, 0.95 for corn, and 0.93 for soybeans, providing robust evidence of the stability and efficiency of the proposed method in generating high-precision crop mapping samples. In-depth discussions and exchanges were held with experts from Japan, Southeast Asia, and Europe and America after the conference, yielding positive outcomes.

Furthermore, during the conference, Dr. Duan Yulin engaged in brief discussions with scholars from the University of Tokyo regarding future international cooperation. They clarified plans to further advance various aspects of exchange cooperation, including collaborative research projects, personnel exchanges, and other mechanisms to expand the breadth and depth of international collaboration.