基于卷积神经网络的SAR舰船检测算法

SAR Ship Detection Algorithm Based on Convolutional Neural Network

合成孔径雷达(SAR)是全天候全天时的传感器,如GF-3、Sentinel-1、TerraSAR-X等,可以产生高分辨率SAR图像,被广泛应用于船舶交通监测、军事和民用领域。SAR图像中的船舶作为重要的军事和民用目标,是需要重点关注的对象。通过针对开放的OpenSARShip数据集进行实验,并引入基于卷积神经网络的检测算法建立相关检测模型,与传统的CFAR算法进行对比分析,结果显示基于卷积神经网络的检测算法可以获得较高的检测精度,要明显高于传统的CFAR检测算法,该研究结果可为SAR图像船舶检测的人工智能化技术提供一些方法上的参考和指导。

Synthetic Aperture Radar(SAR)is an all-day and all-weather sensors,such as GF-3,TerraSAR-X,RADARSAT-2 and Sentinel-1,which can produce high resolution SAR image.SAR images are widely used in ship traffic monitoring,military and civil fields.As an im portant military and civil target,ships in SAR images need to be focused on.In this paper,a detection algorithm based on convolution neu ral network is introduced to establish a detection model,and experiments are carried out on open OpenSARShip data sets.And compared with conventional CFAR algorithms.The results show that the detection algorithm based on convolution neural network can achieve higher detection accuracy,which is significantly higher than the traditional CFAR detection algorithm.The research results of this paper can pro vide some reference and guidance for the artificial intelligent technology of ship detection in SAR images.