An SAR image target detection method based on the convolutional neural network

Abstract

The invention provides an SAR image target detection method based on the convolutional neural network and mainly solves the problem of low detection speed and low detection accuracy rate in the conventional SAR image target detection technology. The method comprises the steps of firstly training a convolutional neural network classification model M0 and designing two to-be-trained convolutional neural network models M1 and M2 based on the model M0; secondly, using an expanded MiniSAR training set as the input for the to-be-trained convolutional neural network models M1 and M2 and acquiring an optimal slice extracting model M12 and an optimal detection model M22 through training; thirdly, detecting the MiniSAR training set by using the optimal slice extracting model M12 and the optimal detection model M22; fourthly, reserving candidate areas with output probability values of the optimal detection model M22 being greater than a threshold value and acquiring a detection result. The method has the advantages of high detection speed and high detection accuracy rate and can be used for vehicle target detection.

Description

technical field [0001] The invention belongs to the technical field of radar, in particular to a target detection method, which can be used for vehicle target detection. Background technique [0002] Since the beginning of the 21st century, convolutional neural networks have been widely used in many fields such as detection, segmentation, and object recognition. In the ImageNet competition in 2012, convolutional neural networks have achieved unprecedented results, which is lower than the best method at that time. Half the error rate, and use the fast graphics processing unit GPU to train the convolutional neural network model. Due to the powerful parallel computing capability of the GPU, the time required for training the convolutional neural network model is greatly shortened. The success of this experiment brings computer vision. Come a new revolution. [0003] Synthetic Aperture Radar (SAR) is an active sensor that uses microwaves for perception. SAR imaging is not limit...

AI Technical Summary

Problems solved by technology

[0006] The present invention expands the MiniSAR training set by using the Mstar data set, and solves the problem that the existing SAR image data cannot satisfy the need for a large amount of training data for the convolutional neural network model. In order to achieve the above invention, the technical solution includes the following:

Images