부산대학교 도서관

The classification study of beetles is of great significance for biodiversity conservation and utilization, agricultural and forestry pest control and other work. In the traditional classification method of beetles, there are problems such as single dimension of extracted features, insufficient feature information, insufficient model accuracy and model generalization, and difficulty in practical application. In order to solve the current problems encountered, this paper proposes an image classification model of Longhorn Beetles of Yunnan based on Bagging and CNN. The model is based on the data set extracted by Gabor filter, constructs LHB-CNN, AlexNet, Xception, InceptionV3, MobileNetV2 to obtain the base classification model. Second, the base model is ensembled through the Bagging ensemble strategy. The experimental results show that the classification model based on Bagging and CNN achieves 99.38% classification accuracy. Compared with the five models of LHB-CNN, AlexNet, Xception, InceptionV3 and MobileNetV2, this method has higher accuracy and precision, and experiments have proved the advantages and better robustness of this method. Therefore, the image recognition model based on Bagging and CNN can effectively classify the images of Longhorn Beetles of Yunnan, and provide certain support for the prevention and control of Longhorn Beetles of Yunnan.