基于神经网络的农作物产量预测方法

Crop Yield Prediction Method Based on Neural Network

准确预测农作物的产量对指导农业生产、保障粮食安全以及维持农业可持续发展具有重要意义。传统的农作物产量预测模型通常对输入数据进行统计和拟合,从而实现农作物产量预测,这些模型一般需要人工选取特征且拟合的方法是线性拟合的,而实际中大多要处理的是非线性问题。研究小组采用深度学习方法,并用已公开的农作物遥感图像数据来训练神经网络。由实验结果可以明显地发现研究小组所提出的神经网络模型识别误差率为5.5左右,优于传统的方法,而且鲁棒性也高于传统方法。

Accurate prediction of crop yield is of great significance to guide agricultural production,ensure food security and maintain sustainable agricultural development.The traditional crop yield prediction model usually makes statistics and fitting on the input data,so as to realize crop yield prediction.These models generally need to select features manually,and the fitting method is linear fitting,but in practice,most of them have to deal with nonlinear problems.Therefore,the research team adopts the deep learning method and trains the neural network with the public crop remote sensing image data.The experimental results show that the recognition error rate of the neural network model proposed by the research team is about 5.5,which is better than the traditional method.And the robustness is also higher than the traditional methods.