Image classification method based on non-subsampled Contourlet transformation and convolutional neural network

Abstract

The invention discloses a convolutional neural network image classification method based on non-subsampled Contourlet transformation. First non-subsampled Contourlet transformation is utilized to makecharacteristic statistics of RGB three-channel images, and relatively good invariance and discrimination description of the images in a transform domain is captured; and then a deep learning method of a convolutional neural network is built on this basis, thereby achieving the purpose of RGB image classification. Learning of known and unknown characteristics is performed through the non-subsampled Contourlet transformation and convolutional neural network, learning of a large number of parameters is avoided, and a subsequent network model can also be simplified.

Description

technical field [0001] The invention relates to the field of image classification, in particular to an image classification method based on non-subsampling Contourlet transformation and convolutional neural network. Background technique [0002] Image classification is to identify and classify images according to the different features reflected in the image objects or scenes. The purpose is to enable the computer to identify the classification of a known image, so as to further understand the image. Image classification technology based on convolutional neural network has been highly valued by researchers due to its high recognition effect. [0003] The main goal of the convolutional neural network is to learn hierarchical image features. The so-called classification refers to the complex function mapping of features from the underlying pixel input to the high-level. Usually, in order to obtain sufficient hierarchical features, it is necessary to increase the learning abil...

AI Technical Summary

Problems solved by technology

Usually, in order to obtain sufficient hierarchical features, it is necessary to increase the learning ability of the convolutional neural network model by increasing the number of hidden layers and the number of neurons, which leads to a lot of parameters to be learned by the neural network, and more parameters will be The training time of the entire network becomes very long, which seriously affects the adjustment of parameters

In addition, too many training samples will also lead to over-fitting of the model.

Images