ass日本风韵熟妇pics男人扒开女人屁屁桶到爽|扒开胸露出奶头亲吻视频|邻居少妇的诱惑|人人妻在线播放|日日摸夜夜摸狠狠摸婷婷|制服 丝袜 人妻|激情熟妇中文字幕|看黄色欧美特一级|日本av人妻系列|高潮对白av,丰满岳妇乱熟妇之荡,日本丰满熟妇乱又伦,日韩欧美一区二区三区在线

Based on the UAV multi-spectral image, the summer corn vegetation index was extracted as a feature variable, and the Pearson correlation coefficient combined with the random forest algorithm was used to reverse the verification weight method for feature selection and redundant features were removed. Random forest, gradient boosting tree, support vector machine and ridge regression were used as the primary learner, and ridge regression was used as the secondary learner to establish a summer corn coverage estimation model based on Stacking ensemble learning, and 5-fold cross-validation was used to further improve model generalization ability, a combination of random search and grid search was used to optimize model hyper parameters, four regression indicators were used for model accuracy evaluation, and the following year’s data was used to verify its robustness. The experimental results showed that compared with a single model and decision tree, Xgboost, Adaboost, and Bagging integrated framework, the Stacking integrated learning model had higher accuracy and stronger robustness. The R2 was 0.9509, which was an average improvement of 0.0369 than that of the single model. Compared with other integrated models, the average increase was 0.0417;RMSE, MAE and MAPE were 0.0432, 0.0330 and 5.01%, respectively, which were 0.0138, 0.0130 and 2.14 percentage points lower than that of the single model, and 0.0185, 0.0126 and 2.15 percentage points lower than that of the other integrated models. The research result provided a method and effective support for the estimation of summer corn coverage.