基于卷积神经网络的洱海湿地昆虫分目识别应用

Application of insects classification in Erhai Wetland based on Convolutional Neural Network

以云南大理洱海湿地为研究背景,针对洱海湿地昆虫形态微小,不易识别的特点,本文基于卷积神经网络,对洱海湿地采集到的6个目的昆虫图像进行了无模板的特征提取及分目识别。首先,对昆虫图片进行预处理;其次,输入到剥除全连接层的VGG16模型中进行特征提取;最后,输入到重构的全连接层进行智能分目识别。实验结果证明,经过迭代训练,训练集的平均识别率达到了95.80%,测试集的平均识别率达到了87.20%,该分目识别算法对昆虫识别有着较高的应用价值。

Taking the Erhai Wetland in Dali, Yunnan Province as the research background, aiming at the characteristics of insects in Erhai Wetland that are small in size and difficult to identify, the template-free feature extraction and identification of six insects images collected in Erhai Wetland based on Convolutional Neural Network are carried out in this paper. In this paper, the insect images are first preprocessed. Then the processed images are put into the VGG16 model stripped of the fully connected layer for feature extraction. Finally, the feature is passed into the reconstructed fully connected layer for intelligent recognition. After iterative training, the average recognition rate of the training set reaches 95.80%, and the average recognition rate of the test set reaches 87.20%. The experimental results show that the recognition algorithm has a high application value for intelligent insects recognition.