Invited Speaker|特邀报告人


Keynote Speaker I 
1636528929711044.jpg  

Dr. Zhuowei Wang   
Australian Artificial Intelligence Institute, University of Technology Sydney, Sydney  


Brief Biography 
He is with the Australian Artificial Intelligence Institute (AAII), University of Technology Sydney. His current research interests include federated learning, noisy label learning, weakly-supervised learning, few-shot learning, image classification, and the corresponding real-world applications.

Title: Predicting rice yield in China with multi-source satellite data from the perspective of artificial intelligence

Abstract: China is the world's largest rice producer, accounting for 28% of global rice production and 41% of China's total grain output. However, rice production has stagnated in recent years. To match projected population growth, China will need to increase rice production by about 20 percent by 2030 to meet domestic demand. Timely, reliable, and large-scale estimates of rice production in China are of great value for policymakers to formulate government food security development plans. Therefore, developing new methods for timely and reliable crop yield estimation over large areas at low cost is necessary.
Machine learning has sophisticated capabilities and capabilities to handle complex relationships between predictors and target variables and can analyze hierarchical and nonlinear relationships between predictors and corresponding variables. In recent years, many studies have developed machine learning-based crop yield prediction models, such as artificial neural networks, LASSO, support vector machine, and random forest. These methods are increasingly used in many countries in agricultural research.
Multi-source environmental data have been used to predict crop yields, such as soil properties and vegetation index, which are important for improving model performance. Satellite remote sensing can directly and timely monitor the growth of crops through various spectral bands.
Aiming at the low-cost and large-scale rice yield prediction problem, we propose a method based on deep learning to predict rice yield. We utilize satellite remote sensing data and machine learning algorithm to complete the segmentation of rice planting areas. We also compare the impact of different classifiers and different data on the segmentation of rice planting areas. Then a rice yield prediction model is designed based on the self-attention mechanism. Our method demonstrates the potential of using NIRv for yield prediction using machine learning. This approach is broadly applicable to other regions globally using publicly available data.



Speaker II
Prof.  Name 
Affiliation 
Title
Abstract


Speaker III
Prof.  Name 
Affiliation 
Title
Abstract


Speaker IV
Prof.  Name 
Affiliation 
Title
Abstract