Polarized SAR terrain classification method based on scattering mechanism multichannel expansion convolutional neural network

Abstract

The invention discloses a polarimetric SAR terrain classification method based on a scattering mechanism multichannel expansion convolutional neural network. The implementation process of the method is as follows: step 1, performing data preprocessing; step 2, performing sample division; step 3, constructing a multichannel convolutional neural network based on a scattering mechanism; step 4, training a network model; step 5, carrying out PolSAR image surface feature classification. According to the method, the complexity in a feature extraction process is reduced, in order to avoid mutual interference among the features, three polarization features are respectively input into the three-channel convolutional neural network according to a scattering mechanism for feature extraction, and the network parameters of each channel are same. Therefore, it is ensured that each channel contributes the same to a final result, and the final classification result also achieves a relatively good effect.

Description

technical field [0001] The invention relates to a polarization SAR image object classification method based on a multi-channel expansion convolutional neural network based on a scattering mechanism, and belongs to the technical field of polarization synthetic aperture radar object classification. Background technique [0002] Full-polarization SAR can work under different transmission and reception polarization combinations, and has the characteristics of higher information content than single-polarization SAR. The measured polarization scattering matrix can completely characterize the polarization scattering characteristics of the target at a specific attitude and observation frequency, and the polarization combination of electromagnetic waves is sensitive to the shape, size, structure, material, etc. of the target, so the full polarization SAR can It has greatly improved the ability to acquire target information, and has been widely used in land cover classification, targe...

AI Technical Summary

Problems solved by technology

[0006] The main purpose of the present invention is to solve the problem of mutual influence between scattering models and feature redundancy in the process of PolSAR image classification, and use dilated convolution instead of ordinary convolution to reduce the information loss caused by downsampling, and provide a scattering mechanism-based PolSAR image object classification method based on multi-channel dilated convolutional neural network (DMCNN)

Images