“ Accurate Predicting Phenological Stages in 327 Rice Varieties “ of Data

Hainan is China's special economic zone and pilot free trade zone, as well as an important national base for Nan Fan. Rice cultivation is a pillar of Hainan's agriculture and a significant source of income for farmers. Rice yield information is a crucial component of precision agriculture and an urgent requirement for high rice yields. Therefore, to grasp the relevant conditions of crop growth stages and physiological and biochemical changes, thus accurately guiding rice field production management to achieve high quality, high yield, and high efficiency production goals, this study explores the ability of single rice plant contour recognition based on deep learning algorithms and the ability to predict the growth stages of multiple rice varieties based on vegetation indices.This study first constructs a multi-variety single rice plant contour recognition model based on Cascade R-CNN to explore the generalization ability of deep learning algorithms in single plant contour recognition of different rice varieties. Then, a growth stage prediction model for 327 rice varieties is established based on NDVI to explore the accuracy and feasibility of remote sensing prediction of rice growth stages. This study is located in the National Rice Park of Haitang Bay, Sanya City, Hainan Province. The geographical coordinates are 109°68′22″41E-18°31′97″N, and the climate conditions are tropical marine monsoon climate with an annual average temperature of 25.7℃, annual sunshine duration of 2534 hours, and an annual average precipitation of 1347.5 mm. The original data was collected during different growth stages: greening stage on August 16, 2023, with a maximum temperature of 34℃ and a minimum temperature of 27℃; tillering stage on September 1, 2023, with a maximum temperature of 32℃ and a minimum temperature of 26℃; jointing stage on September 18, 2023, with a maximum temperature of 34℃ and a minimum temperature of 25℃; heading stage on October 16 and October 26, 2023, with temperatures of maximum 31℃, minimum 25℃, and maximum 30℃, minimum 25℃ respectively; and mature stage on November 6 and November 13, 2023, with temperatures of maximum 29℃, minimum 24℃, and maximum 28℃, minimum 24℃ respectively. A total of 5 growth stages and 327 rice planting areas were monitored. However, due to the limitation on the size of data upload, only the data from the tillering stage was uploaded. Below is the description of our data package: The data in folder (Data-01) includes orthoimages of the research area. Remote sensing images of each growth stage were obtained using DJI Mavic 3M drone, set at a flight altitude of 30m, takeoff speed of 10 m/s, flight path speed of 2.6 m/s, with overlap settings of 70% front overlap and 80% side overlap, and the shooting mode was set to interval shooting. The data in folder (Data-02) includes NDVI index maps of the research area. The original images obtained after shooting were uploaded and processed using DJI Terra for image correction and alignment. The index calculation tool was used to select the required index type, set the corresponding band combinations, and calculate the NDVI index images for the greening, tillering, jointing, heading, and mature stages. The data in folder (Data-03) includes the results of 327 rice varieties identified by the MMDetection Object Detection (Cascade R-CNN) model for different rice varieties. The data in folder (Data-04) includes the location of each rice plant identified by the deep learning model for 327 rice varieties, as well as the NDVI values at different growth stages. The "OBJECTID" in each table represents the number of rice plants in the image, "x" represents the longitude of each rice plant, "y" represents the latitude of each rice plant, and "NDVI (20230816-1113)" represents the NDVI values of rice plants at different growth stages. If you need high-resolution raster format raw data for other periods, please contact Dr. Zixuan Qiu at [e-mail].