基于卷积神经网络的白背飞虱识别方法

为了实现白背飞虱虫情信息的自动收集和监测,提出一种基于卷积神经网络的白背飞虱识别方法并进行应用研究。首先,用改进的野外环境昆虫图像自动采集装置,采集田间自然状态下的白背飞虱图像,对所获取的图像进行归一化处理。然后,随机选取1/2图像样本作为训练集、1/4作为测试集。利用5×5卷积核对训练样本进行卷积操作,将所获取的特征图以2×2邻域进行池化操作。再次经过卷积操作和3×3邻域池化操作后,通过自动学习获取网络模型参数和确定网络模型参数,得到白背飞虱的最佳网络识别模型。试验结果显示,利用训练后的网络识别模型,对训练集白背飞虱的识别正确率可达96.17%,对测试集白背飞虱的识别正确率为94.14%。

Automatic Identification Method for Sogatella furcifera Based on Convolutional Neural Network

In order to realize the pest information automatic collection and monitoring for Sogatella furcifera, an automatic recognition method based on convolutional neural network was presented and its application was carried out. The images of Sogatella furcifera were collected in the natural state of the field by using the improved automatic acquisition system for insect images in field environment and the acquired images were normalized. Six hundred of images were randomly selected from the normalized images as training set and three hundred ones were chosen as test set. The convolution operation was performed on the training set with 5×5 convolution kernel and the acquired feature graphs were pooled in a 2×2 neighborhood. After the convolution operation and 3×3 neighborhood pooling operation, the network model parameters were obtained by using automatic learning and the optimal network identification model for Sogatella furcifera was achieved. The experimental results showed that the recognition accuracy for Sogatella furcifera could reach 96.17% for training set, and for test set, the recognition accuracy was 94.14%.