Construction of general platform for agricultural diseases diagnosis based on fuzzy neural network | 基于模糊神经网络的农业病害诊断通用平台构建

中国人. 构建基于模糊神经网络的农业病害诊断通用平台。将客观计算与专家经验相结合，分别对病害症状的重要性及其模糊描述进行分析量化。建立模糊BP神经网络通用诊断模型用于诊断推理，引入附加动量法和自适应学习速率法对传统BP算法进行改进。采用面向对象的方法，运用Java语言开发病害动态诊断平台，并以葡萄病害诊断为例，对平台功能及性能进行测试。基于模糊神经网络的农业病害诊断通用平台构建成功，其对葡萄病害的诊断结果与专一的葡萄病害诊断专家系统诊断结果一致，且神经网络的训练速度显著提高。构建的基于模糊神经网络的农业病害诊断通用平台,适用于多种作物的病害诊断，克服了完全依靠专家经验的主观性，诊断效率高，具有较高的实用性、通用性和灵活性。

英语. The purpose of the present work was to construct a general platform for agriculture diseases diagnosis based on fuzzy neural network. The importance and fuzzy description of the symptoms of diseases were analyzed and quantified based on both objective computation and expert experience．In the fuzzy back-propagation, neural network general diagnosis model, which was used in diagnostic reasoning, additional momentum method and adaptive learning rate method were introduced in order to improve the traditional BP algorithm. The development of platform used Java technology, an Object-Oriented method. Take grape diseases diagnosis for example, the function and performance of platform were tested. The construction of general platform for agricultural diseases diagnosis based on fuzzy neural network was successful. The diagnosis results of the platform constructed in this paper and single grape diseases diagnosis expert system were the same, and the training speed of neural network was greatly improved. The constructed general platform was applicable to disease diagnosis for multiple crop, which overcame the subjectivity of being diagnosed by expert completely. The operation efficiency of platform was improved too and it had higher practicability, generality and flexibility.