The largescale crop yield forecasting is of great significance to grasp the state of national grain production timely and accurately and carry out effective grain macrocontrol. To improve the timeliness of maize yield forecasting, taking central plain of Hebei Province as the study area, yield forecasting was carried out for period during 2016 to 2018. The Savitzky-Golay filtered leaf area index, closely related to maize growth and yield, was selected as the characteristic parameter. The LAI data from early July 2010 to late August 2018 were used as the modeling data, and the LAI data from early September to late September of each year from 2016 to 2018 were used as the test data. The LAI data were extracted pixel by pixel to form a onedimensional time series as the input data of the model. Based on the autoregressive integrated moving average (ARIMA) model and radial basis function (RBF) neural network, LAI data of the study area were forecasted pixel by pixel. And the average absolute error and root mean square error were used to evaluate the prediction accuracy of the two models. The results showed that the accuracy of LAI forecasting based on the ARIMA model was better than that of RBF neural network. The RMSE of step1 and step2 LAI forecasting results was 0.18m2/m2 and 0.14m2/m2 respectively, which was lower than that of RBF neural network, indicating that the ARIMA model was more suitable for forecasting summer maize yield per unit area in the central plain of Hebei Province. Based on the research correlation of weighted LAI and summer maize yield, and ARIMA LAI forecasting results, the summer maize yield forecasting models were developed at intervals of 1ten day, 2ten day, 3ten day before the harvest. The results showed that the forecasting accuracy of maize yield per unit area at 1ten day, 2ten day, 3ten day intervals was high in both the county scale and pixel scale, and the maximum relative error between the forecasting and monitoring of yield per unit area in the county (district) scale from 2016 to 2018 was only 373%. The method can be used to forecast summer maize yield at 30 days before the harvest.