基于卷积神经网络的红木识别App设计与开发

Design and Development of Redwood Recognition App Based on Convolutional Neural Network

为了实现对红木快速自动识别,笔者设计和开发了一款基于卷积神经网络的红木识别App。首先利用无损显微设备获取木材横切面和弦切面的微观构造图像,将其作为训练数据;其次在TensorFlow框架上构造MobileNet网络,创建数据集并进行训练,以得到红木识别模型;最后在此基础上设计和开发出红木识别App。所开发的App具有占用手机内存小、页面简洁明了等特点,方便用户通过手机实现红木类型的自动识别。

In order to realize fast and automatic identification of mahogany,the author designed and developed a mahogany identification app based on convolutional neural network.Firstly,the microscopic structural images of the wood cross-section and chord section are obtained by non-destructive microscopy equipment,and used as training data;Secondly,the MobileNet network is constructed on the TensorFlow framework,and the data set is created and trained to obtain the redwood recognition model;Finally,on this basis design and develop a redwood identification App.The developed App has the characteristics of small occupation of mobile phone memory,concise and clear pages,et al,which is convenient for users to realize automatic identification of mahogany types through mobile phones.