基于深度学习算法的农作物灾害预测研究

Research on crop disease prediction based on deep learning algorithm

针对人工预测农作物病害的方法存在效率低、误差大的弊端,提出一种基于深度学习的农作物病害预测方法。首先采用基于改进深度学习的特征提取算法,提取农作物特征;基于提取的农作物特征,再通过基于粒子群支持向量机状态识别的农作物病害识别模型,实现农作物病害预测。实验结果表明:所提方法对农作物特征提取耗时最大值为54.76 ms,提取精度最大值为0.983;对同一农作物不同病害预测精度高达0.94,对不同农作物的同一病害、不同农作物差异病害的预测误差值均为0.02。某农科院采用该方法对马铃薯病害进行预测后,预测效果的满意态度达到100%,由此验证所提方法对农作物病害预测具有一定应用价值,预测性能显著。

In allusion to the disadvantages of low efficiency and large error in the artificial prediction of crop diseases,a method of crop disease detection based on deep learning is proposed.The feature extraction algorithm based on improved deep learning is used to extract the features of crops.Based on the extracted crops features,the crop disease recognition model based on the state recognition of particle swarm support vector machine is used to realize the prediction of crop diseases.The experimental results show that the maximum time⁃consumption and maximum extraction precision value of the proposed method is 54.76 ms and 0.983 respectively,the prediction precision of different diseases of the same crop is as high as 0.94,and the prediction error of the same disease and different diseases of different crops is 0.02.After a certain academy of agricultural sciences used this method to predict potato diseases,the satisfactory attitude of prediction effect reaches 100%,which verifies that the proposed method has a certain application value for crop disease prediction,and the prediction performance is significant.