SAR image target detection method based on fused convolutional neural network

Abstract

The invention discloses an SAR image target detection method based on a fused convolutional neural network, and the method comprises the steps: enabling a detection model to be better converged, training a classification model through a relatively complete classification data set, and carrying out the parameter initialization of a slice selection model and the detection model through the model; the method comprises the following specific steps: (1) designing a convolutional neural network classification model based on existing classification data; (1a) building a convolutional neural network; and (1b) setting convolutional neural network parameters. The SAR image target detection method has the advantages that the convolutional neural network SAR image target detection method using the extended sample to assist model training is provided, and the extended sample is used to train a parameter initialization model; proper size transformation processing is conducted on the expanded sample, a training data set is formed by combining the expanded sample with a target domain sample, and a slice selection model and a detection model capable of sharing all convolution layers are obtained by utilizing a loop training method; an auxiliary training strategy can help the model to better converge, and the algorithm is effective.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for detecting a SAR image target based on a fusion convolutional neural network. Background technique [0002] Convolutional Neural Networks (CNN) is a type of feed-forward neural network that includes convolution calculations and has a deep structure. It is one of the representative algorithms for deep learning. The convolutional neural network has the ability to learn representations, and can perform translation-invariant classification of input information according to its hierarchical structure, so it is also called "translation-invariant artificial neural network". [0003] SAR (Synthetic Aperture Radar), that is, synthetic aperture radar, is an active earth observation system that can be installed on aircraft, satellites, spacecraft and other flight platforms to observe the earth all-weather and all-weather, and has a certain surface penetration capabilit...

AI Technical Summary

Problems solved by technology

[0006] According to the Faster R-CNN detection model to complete target detection on complex SAR images, there are problems of lack of training data and difficulty in model training

Images