Image Classification Method Based on Sparse Autoencoder and Support Vector Machine

Abstract

The image classification method based on sparse autoencoder and support vector machine includes the following steps: obtain training image set and test image set; construct a deep sparse autoencoder with multiple hidden layers, and train the deep sparse autoencoder until it satisfies training conditions. The test image set is input to the trained deep sparse autoencoder for layer-by-layer learning; and the proposed feature weight combination method is used to assign a feature weight to each feature set, and finally the feature set obtained by each hidden layer and The corresponding weights are combined into a new image feature set to be classified; the training image feature set is used as a training sample to train the support vector machine classifier to achieve the optimal classification performance of the support vector machine; the obtained image feature set to be classified is input To the optimized support vector machine, image classification is performed and the classification accuracy is obtained. The invention effectively improves the classification accuracy of images, and has good generalization ability and applicable scope.

Description

technical field [0001] The invention relates to an image classification method, which belongs to the technical fields of pattern recognition, intelligent calculation and image processing, and in particular to an image classification method based on a sparse automatic encoder and a support vector machine. Background technique [0002] With the development of information technology, image data has increased dramatically, and the demand for image processing has also greatly increased. In real life, due to factors such as blurred images, unclear fonts, and shooting angles, the quality of collected images is often not high. affect the accuracy of image classification. Traditional image classification techniques are mostly based on large-scale calculations, and there is often a certain contradiction between their calculation volume and calculation accuracy. In the early image classification research, the underlying features of the image were studied, and low-level features such a...

AI Technical Summary

Problems solved by technology

[0004] The shortcomings of the existing methods: On the one hand, it is difficult for the single-layer sparse autoencoder to extract complete and deep abstract features during feature learning, so part of the feature information will be lost, making the feature lack of robustness, which in turn affects the performance of image classification. Accuracy; on the other hand: the support vector machine is affected by factors such as parameters and kernel functions, and the performance of the support vector machine will affect the final image classification accuracy

Images