Vegetables are one of the main crops in China, and pests are one of the important factors affecting the quality of vegetables. In order to improve the recognition accuracy of vegetable pest images, a vegetable pest image recognition method based on improved VGG convolution neural network is proposed. Based on the VGG16 and VGG19 models, the method optimizes the number of full connection layers, replaces the original SoftMax classifier in VGGNet with the three-label SoftMax classifier, optimizes the structure and parameters of the model, and uses the weight parameters of convolution layer and pooling layer in the pre-training model in transfer learning. Experiments were carried out on the self-expanding data set of vegetable pest images, and the performance of the method was tested. Tensorflow was used to train the network model. The experimental results showed that the pre-trained models (VGG16, VGG19, Inception V3, ResNet50) were trained on the vegetable pest image data set to adapt to the recognition of vegetable pest images. The experimental results also showed that compared with Inception V3 and ResNet50, the recognition accuracy of the pre-trained models using VGG16 and VGG19 were higher, and the test accuracy of the two models were 99.90% and 99.99% respectively. Finally, the methods were compared with the traditional VGG method in self-expanding data sets. The results showed that the accuracy of VGG16 model and VGG19 model were improved from 85.90% and 86.21% to 100% and 100% respectively; the classification accuracy of VGG16 model was improved from 64.02% to 99.90%, and the classification accuracy of VGG19 model was improved from 85.83% to 99.99%, which effectively improved the recognition accuracy.

No abstract

Concept modeling and learning have been important research topics in artificial intelligence and knowledge discovery. This paper studies a hierarchical concept learning method that requires a small amount of data to achieve competitive performances. The method starts from a set of fuzzy prototypes called Fuzzy Semantic Cells (FSCs). As a result of FSC parameter optimization, it creates a hierarchical structure of data–prototype–concept. Experiments are conducted to demonstrate the effectiveness of our approach in a classification problem. In particular, when faced with limited training data, our proposed method is comparable with traditional techniques in terms of robustness and generalization ability.

No abstract