Abstract: Typical japonica rice in the northeast area was taken as an example, the japonica rice leaf vegetation index NDVI and PRI were measured by using vegetation index measuring instrument PlantPen, and the growth process of rice was divided into four growth periods in accord with the phenological process. Firstly, the correlation analysis of NDVI and PRI was carried out by using the method of dual distance variable correlation analysis; Then, the NDVI fitting regression model of PRI was established by using linear regression and Cubic curve regression, and the goodness of fit and accuracy of regression model were verified; meanwhile, the fitting effect and test results of linear regression model with Cubic curve regression model were analyzed. The results showed that the leaf vegetation index NDVI and PRI in different growth periods of japonica rice showed significant correlation, and the correlation increased with the growth process of japonica rice. Both of the linear regression model and Cubic curve regression model could make good fitting PRI, NDVI in japonica rice growth process, and the fitting effect became better and better. The four corresponding indexes determination coefficient(R²), root mean square error (RMSE), absolute percentage error (MAPE) of Cubic curve regression model were 0.805 5, 0.035 8, 0.534%; and those of the linear regression model were 0.765 3, 0.048 8, 1.365%. It was obvious that the Cubic curve regression model had smaller RMSE and MAPE values and larger R² value. Thus, its goodness of fit and inspection accuracy were better than the simple linear regression model, which could be used as a reference model for NDVI inversion PRI.

Key words: japonica rice, NDVI, correlation, regression analysis