Polarized SAR image classification method and system based on federal evolution convolutional neural network

Abstract

The invention discloses a polarized SAR image classification method and system based on a federated evolution convolutional neural network. The method comprises the following steps: dividing polarized SAR data into a training set and a verification set; decoding individuals in the population into a convolutional neural network, and inputting the convolutional neural network into a training set for training and aggregation; repeatedly operating the aggregated convolutional neural network to obtain a final aggregated convolutional neural network; inputting the verification set into the final aggregated convolutional neural network to obtain the number of correctly classified samples of the verification set; performing differential evolution operation on each individual in the population to obtain a filial generation population; combining the population with the offspring population to obtain a combined population; executing environment selection operation to obtain a next-generation population; selecting a final population and selecting the convolutional neural network corresponding to the individual with the highest fitness from the final population; and inputting the to-be-classified polarimetric SAR image into the convolutional neural network to obtain a classification result, and completing polarimetric SAR image classification. According to the method, the classification precision of the convolutional neural network is improved, and meanwhile, the complexity of the convolutional neural network is reduced.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a polarization SAR image classification method and system based on a federated evolutionary convolutional neural network. Background technique [0002] Polarimetric SAR image classification is an important step in the application of polarimetric SAR, and it is also a classic research direction of polarimetric SAR image processing. At present, polarimetric SAR has great application value in both civilian and military fields, and more and more institutions have begun to establish polarimetric SAR databases. However, due to data privacy and other factors, the polarization SAR data among various institutions cannot be fully utilized. Therefore, it is urgent to propose a method that can fully mine the value of polarization SAR data under the premise of protecting data privacy. [0003] A polarimetric SAR image classification method based on channel attention deep...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a polarimetric SAR image classification method and system based on federated evolutionary convolutional neural network in view of the deficiencies in the above-mentioned prior art, so as to solve the problem that the prior art cannot protect data privacy. The problem of automatically designing high-precision and high-efficiency neural network models by combining multiple data sources

Images