Radar image spot noise suppression method based on correlation loss convolutional neural network

Abstract

The invention discloses a radar image spot noise suppression method based on the correlation loss convolutional neural network. The method comprises steps that a radar image is inputted into the trained convolutional neural network to obtain the spot noise suppression result image; the training method of the convolutional neural network includes steps that spot noise is superimposed to a remote sensing image to obtain the image after spot noise superimposition, and the image after spot noise superimposition is inputted into the convolutional neural network to obtain a training output image; error loss is obtained according to a mean square error of the remote sensing image and the training output image, and correlation loss is obtained through utilizing a correlation coefficient between the remote sensing image and the training output image; reverse propagation is performed through utilizing the error loss and the correlation loss, and weight parameters of the convolutional neural network are updated; the trained convolutional neural network is obtained. The method is advantaged in that the error loss and the correlation loss are utilized, the weight parameters of the convolutionalneural network are updated, and thereby the radar image spot noise suppression effect of the network is effectively improved.

Description

technical field [0001] The invention belongs to the field of radar image processing, and more specifically relates to a radar image speckle noise suppression method based on a correlation loss convolutional neural network. Background technique [0002] Radar plays a very important role in many fields of military and civilian use. Moreover, with the continuous upgrading of radar imaging technology and the improvement of radar image resolution, its application range is also expanding. [0003] However, due to its inherent radar imaging mechanism, radar images usually have speckle noise. Speckle noise is usually a kind of multiplicative noise, which is caused by mutual interference of echo signals scattered by several reflection points in each resolution unit. Speckle noise has caused great troubles to subsequent processing and information extraction, greatly affecting the practical performance of radar images. [0004] At present, in the field of radar image speckle noise su...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above defects or improvement needs of the prior art, the present invention provides a radar image speckle noise suppression method based on a correlation loss convolutional neural network, thus solving the lack of self-adaptability of the prior art and the poor expressiveness of the model. Weak, unable to effectively suppress the technical problem of speckle noise

Images