基于VGG-16卷积神经网络的水稻疾病识别小程序

Rice Disease Recognition Wechat Applets Based on VGG-16 Convolutional Neural Network

微信小程序用户群体多、范围广、操作灵活便捷等优点得到广泛使用,其“用完即走”的特点非常适合于解决农民对常见水稻虫害识别不熟悉的而导致农药滥用错用问题。为了实现自然场景下害虫实时精准被识别,构建基于VGG-16卷积神经网络的水稻害虫智能识别模型。根据水稻的叶片特征和发病自然场景,对VGG-16网络的卷积层局部调整,优化主要模型参数,实现针对水稻疾病的智能识别,在测试集上的平均准确率是91.3%,该小程序通过微信开发者工具进行编译,应用WXML、WXSS和Java Script技术,VGG卷积神经网络模型导入常见水稻虫害的数据集训练针对水稻的虫害识别模型,通过Https请求来调用远程服务器运行的模型应用,从而实现小程序识别多种常见水稻疾病的功能。该小程序平台能够帮助用户方便快捷地查询到水稻虫害类别,代替以往传统的农学专家人工辨认的过程,能够快速地进行根据农作物疾病类型进行对应的处理,有效提高水稻病害的防治率,实现精准防治的目标。

Wechat applet has many users,a wide range,flexible and convenient operation and other advantages,and its"run out"feature is very suitable for solving the problem of pesticide abuse and misuse caused by farmers'unfamiliar identification of common rice pests.In order to realize the real-time and accurate identification of pests in natural scenes,an intelligent identification model of rice pests based on vgg-16 convolutional neural network was constructed.According to the leaf characteristics of rice and the natural scene of disease onset,the convolution layer of vgg-16 network is locally adjusted to optimize the main model parameters.The applet is compiled through wechat developer tools,and wxml,wxss and JavaScript technologies are applied.The VGG convolution neural network model is imported into the data set of common rice pests to train the pest identification model for rice,and the model application run by the remote server is called through HTTPS requests,So as to realize the function of the applet to identify a variety of common rice diseases.This small program can quickly carry out the corresponding treatment according to the type of crop diseases,effectively improve the prevention and control rate of rice diseases,and achieve the goal of precise prevention and control.