Biomass data and prediction model of rice heading stage in Haitang district, Sanya City, Hainan Province.

In the context of the rapid development of agricultural science and technology, precision agriculture technology has become a key means to improve crop yield and quality. The aim of this study was to explore the biomass monitoring technology of rice growth in order to provide technical support for the scientific management of rice production in China. In this context, on September 5, 2023, our research group carried out a special study on rice heading biomass for the rice sample field in Haitang District, Sanya City, Hainan Province. A total of 20 representative rice plants at heading stage were obtained in this sample. In order to record the growth of rice comprehensively, this study used UAV technology to collect multi-spectral image of paddy field before sampling, and accurately recorded the geographical coordinate information of the sample. After sampling, the rice plants were dried in an oven at 80℃ for 30 minutes, and then continued to be dried at 120℃ to constant weight to obtain the biomass data of rice heading stage. In the process of multispectral image analysis, the rice plants were labeled by accurate positioning technology, and the corresponding multispectral average value was extracted to reflect the growth state of the plants. On this basis, five biomass characteristic factors OSAVI, Green, Red Edge, Mean_Green and Entropy_Red Edge were selected in this study, with positive correlation above 0.5 and VIF less than 10. In this study, MATLAB R2024a software was used to model and analyze the biomass of rice at heading stage. After rigorous model screening and validation, the Back Propagation(BP) model is finally identified as the best algorithm model, whose validation set R² value is 0.689, root mean square error (RMSE) is 3.956Mg/ha, and relative Root mean square error (RRMSE) is 22.050%. The results show that the model has high prediction accuracy and reliability.The following is the official description of the data folder for this study:1. "Drone Multi-spectral Image" folder: contains vegetation index, band and texture feature images of rice heading stage, providing detailed spectral data for the study.2. "Data Set" folder: contains all data extracted from UAV images of sampling points, raw biomass modeling data and normalized biomass modeling data, providing rich data resources for subsequent research.3. "Model" folder: The MATLAB R2024a rice heading stage modeling code and prediction code are stored, providing practical research tools for researchers in related fields.