Big Data Visualization Analysis of Rice Seedling Quantity Based on UAV

In order to reduce the labor cost and improve the efficiency of rice basic seedling statistics, a statistical method based on computer vision was designed. This paper introduced big data of rice spectrum and mixed pixels, established a basic seedling number inversion model, and drew a basic seedling density map, which effectively improved the statistical accuracy of the basic seedling number of rice, saved the labor cost, and intuitively experienced the basic seedling density of rice, and guided the fertilization and nitrogen application work in the later stage according to the requirements. To achieve the purpose of saving fertilizer and increasing yield in the process of rice planting. In this study, two methods were used to retrieve the number of basic rice seedlings. The first method was based on the pixel binary classification of Normalized Vegetation index (NDVI), visible Vegetation index (VDVI) and green Normalized Vegetation Index (GNDVI), with the coefficient of determination (R
²
) being 0.782, 0.767 and 0.744, respectively. The root mean square errors (RMSE) were 12.44, 12.85 and 13.48, respectively. The other method is a hybrid pixel decomposition method based on fully constrained least squares with an R
²
of 0.897 and an RMSE of 8.53. Comprehensive analysis shows that the seedling number inversion model based on hybrid pixel decomposition has higher accuracy