In this paper, problems common nitrogen fertilizer used in excess of the current agricultural beet production, the establishment of critical real-time accurate nitrogen fertilizer recommendation system.In this paper, by using the BP neural network algorithm using image data to predict nitrogen content of beet, through reasonable excluding abnormal data from the original image data does not match the shooting conditions selected 147 sets of data as a training set, 90 sets of data for the prediction set group.The R, G, B as an input to obtain the predicted value and the actual value of the BP neural network algorithm trained by the best correlation coefficient of r =0 .70 , root mean square error RMSE =4 .60 . The R /(R +G +B), G /(R +G +B), B /(R +G +B) as an input, the use of BP neural network algorithm trained after the predicted value and the actual value of the best correlation coefficient r =0 .64 , root mean square error RMSE =3.66.As can be seen, the use of BP neural network algorithm for establishing beet color feature information ni-trogen model is feasible, provide methodological support for agricultural production in real-time lossless diagnostic beet plant nitrogen content.%当前农业甜菜生产中存在普遍的氮肥过量使用的问题，建立实时准确农田氮肥推荐体系至关重要。为此，通过利用BP神经网络算法利用图像数据对甜菜氮素含量进行预测，通过合理剔除原始数据中不符合拍摄条件的异常图像数据，选取147组数据作为训练集，90组数据组为预测集，将 R、G、B 作为输入量，通过 BP 神经网络算法训练得到预测值与实际值最优相关系数为r＝0．70，均方根误差RMSE＝4．60。将R／（R＋G＋B）、G／（R＋G＋B）、B／（R＋G＋B）作为输入量，利用BP神经网络算法训练后预测值与实际值最优相关系数 r＝0．64，均方根误差RMSE＝3．66。由此可以看出：使用 BP神经网络算法建立甜菜颜色特征信息氮素模型是可行的，可为农业甜菜生产中实时无损诊断植株氮素含量提供方法支持。
展开▼
