Author(s)

Fang Meng, Guogen Fan

Corresponding Author

Fang Meng

ABSTRACT

Based on the traditional fuzzy BP classification method, the image with high degree of feature phase was classified with higher misclassification rate. Considering the problems of the traditional methods, in this paper, a classification feature similarity image classification algorithm based on deep learning and support vector machine was proposed. Firstly, the local average noise reduction method was used to denoise the similarity image, and the wavelet image was decomposed by wavelet multi-scale decomposition algorithm. Then, the local information smoothing processing of the image was performed by the RGB color component recombination method, and the rough set feature quantity of the image was extracted. Finally, the extracted feature quantities were input into a support vector machine learner for image classification. In the hidden layer of the classifier, adaptive learning of weighting parameters was performed by the deep learning algorithm to achieve image enhancement processing and classification optimization of batch feature similarity images. The simulation results showed that the accuracy of feature similarity image classification was better, the ability to resist inter-class attribute perturbation was stronger, and the retrieval efficiency of large-scale images was improved.

KEYWORDS

Similarity image, Classification algorithm