基于多层感知器网络的农作物疾病诊断系统  

Study of Diagnosis for Diseases of Agricultural Crops Based on Multi-Layer Perceptrom Networks

作　　者：陈晓艳[1] 董朝轶[2] 李永亭[1] 刘月文[1] 

机构地区：[1]内蒙古工业大学电力学院电工基础教学中心,内蒙古自治区呼和浩特010080 [2]内蒙古工业大学电力学院自动化系,内蒙古自治区呼和浩特010080

出　　处：《计算机技术与发展》2011年第11期84-88,共5页Computer Technology and Development

基　　金：2010年度内蒙古自治区教育自然科学研究重点项目(NJ10070);内蒙古工业大学校基金(X200416X200805)

摘　　要：农作物疾病的人工诊断效果常受到个人诊断经验和能力的限制,无法达到最令人满意的诊断结果。将丰富的植物病理学诊断经验和知识编入专家系统,利用模式识别算法,对农作物常见疾病进行诊断,可以大大提高诊断准确率,有效地提高其产量和质量。主要研究了基于一种人工神经元网络—多层感知器网络的模式识别技术在大豆疾病诊断中的应用。MLP神经网络通过模拟生物神经元细胞对外部刺激而产生的反应,构成一种前向神经网络,可以有效地解决非线性不可分问题。首先对大豆常见19种疾病症状进行了收集和整理,构建试验样本集。然后,利用反向传播算法对该网络进行训练和测试。测试结果表明,该模型具有较高的农作物疾病诊断正确率和良好的泛化能力。The manual diagnosis for the diseases of agricultural crops is often restricted by the individual ability and experiences so that one cannot obtain the precise results of diagnosis. To overcome this pitfall, the merge of expert systems with the rich pathological knowledge and the utilization of pattern recognition algorithm can significantly improve the precision of diagnosis. Therefore, it greatiy increases the quantity and the quality of crop＇ s production. A pattern recognition technique based on a multi-layer perceptron （MLP） neural network is applied to the diagnosis of soybean diseases. The MLP neural network, which is a novel and efficient feed-forward network, is based on the reflections of cortical neurons on the external stimulus and can be used to solve the problems of nonlinear inseparability. Firstly, nineteen typical symptoms of soybean diseases are collected, and then constructed to form an experimental sample set. Secondly, set up and train the MLP network model using back propagation algorithm. Finally, the test shows that the RBF model has high diagnosis precision and strong generalization ability.

关 键 词：反向传播算法 多层感知器网络 疾病诊断 模式识别 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] S435.2[自动化与计算机技术—控制科学与工程]