【作者】 郑晓洁；

【导师】 李书琴；

【作者基本信息】 西北农林科技大学 ， 计算机应用技术， 2011， 硕士

【摘要】 苹果作为农民经济的来源,苹果树腐烂病的发生直接影响和制约着苹果的品质和产量。传统苹果树腐烂病的预测方法带有主观性,主要通过专家和果农长期的田间经验而对苹果树腐烂病进行预测。为了对苹果树腐烂病进行更加准确科学的预测,设计开发了苹果树腐烂病预测模型,更好的指导农民预防腐烂病的发生。本文首先对影响苹果树腐烂病的因素进行主成分分析,在此基础上,建立了基于BP神经网络的预测模型和小波网络的预测模型,完成了苹果树腐烂病预测模型。本文主要研究内容如下。(1)建立基于BP神经网络预测模型。首先对预测模型的输入端数据进行预处理,然后对模型进行训练,最后通过实验来验证该模型的预测精度。对影响苹果树腐烂病的因素进行了分析和总结,通过主成分分析法提取影响苹果树腐烂的主因素,进行降维。作为BP神经网络的输入,以流行程度作为输出,进行模型的训练和预测。使用洛川县气象站和植保站提供的数据对模型进行实验,结果表明,该模型的预测精度良好。(2)建立基于小波网络的预测模型。分析了“紧致性”和“松散性”小波网络各自优点和不足,提出将“紧致型”小波网络应用于苹果树腐烂病流行程度的预测。通过主成分分析法对历史数据进行分析,利用得到的主成分作为小波网络的输入,选用Morlet小波函数作为隐层激活函数,对洛川县苹果树腐烂病某一年的流行程度进行预测。实验结果证明,该模型预测精度高。(3)构建苹果树腐烂病预测模型。苹果树腐烂病预测模型是基于两种不同的预测模型。用户可以根据当地实际情况选择不同的预测模型进行准确、科学的预测。而且用户也可以输入当地的苹果树腐烂病信息,对苹果树腐烂病进行预测。对当地果农及时做好防治准备工作具有重要意义,而且有利于果农根据预测流行程度的严重性较早制订防治腐烂病的方案,减少经济损失。更多还原

【Abstract】 Apples are important economic crops, and apple tree canker not only affect and restrict the production of apples, but also has a great impact on the quality of apples. The traditional apple canker forecast is based on the experience of farmers and experts on the basis of subjective。In order to establish a more scientific and effective the prediction model of prevalence of apple canker disease for better guiding the farmers.This paper, firstly, by analysis mainly factors influencing apple canker disease based on Principal Component Analysis(PCA),establishing the prediction model of Error Back Proragation(BP)neural network and the prediction model of Wavelet Neural Network(WNN), according to the first step.so as to complete the prediction model of apple tree canker.The mainly research in this paper are the following:(1) BP prediction model. Firstly, input data of the prediction model is treatmented, then the model training and prediction, finally, to verify the prediction accuracy by experiments. Affecting factors of Apple Tree Canker are analyzed and summarized, the extracted main factors by using PCA, as the inputs of the BP prediction model, the morbidity levels of the apple tree canker are used as the outputs. The model show that the prediction accuracy is good.(2)WNN prediction model. The construction of Compaction and Decomposition wavelet neural network are analyzed, a wavelet neural network of compaction construction is used in this paper. the extracted main factors by using PCA, as the inputs of the WNN, the hidden layer activation function used Morlet wavelet function, to prediction morbidity levels of the apple tree canker in luochuan county. At last, the prediction WNN model is based on wavelet transform. The model show that the prediction accuracy is better than the BP prediction model.(3)The Model of Apple Tree Canker. This model provide more the information of disease to users. By comparison two different prediction models, users can choose different model for scoemtific and accurate prediction according to different areas,which is an important role in preparation works of prevention disease to local farmmers.更多还原