一种改进人类视觉的SAR图像舰船检测方法

针对以离散余弦变换为核心的人类视觉模型舰船检测算法受数据类型限制的问题(即对复数类型的数据检测效果不好),该文提出了一种改进的人类视觉模型SAR图像舰船检测算法。该算法是以快速傅里叶变换代替离散余弦变换,将SAR图像从空间域变换到频率域;快速傅里叶变换对数据类型要求较低,只要求数据是离散的,并且运行效率更高。然后,采用3种星载SAR数据——ENVISAT ASAR(25m)、Sentinel-1(10m)和Cosmo-Skymed(2.5m)进行对比实验。结果表明,以快速傅里叶变换为核心的人类视觉模型舰船检测算法的检测性能和效率优于以离散余弦变换为核心的算法、双参数恒虚警率(CFAR)算法和K分布恒虚警率算法。

A method for ship detection in SAR imagery based on improved human visual attention system

Since discrete Cosine transform as the core of the human visual model of ship detection algorithm is affected by the limitation of the data type(i.e.the detection effect of the complex type of data is not good),an improved human visual model of ship detection algorithm for SAR image was proposed in this paper.This algorithm is based on the fast Fourier transform instead of the discrete Cosine transform,which transformed the SAR image from the spatial domain to the frequency domain.The request of the data type of fast Fourier transform is low,it only requires the data is discrete,and the efficiency is higher.Then,three kinds of space-borne SAR data-ENVISAT ASAR(25 m)、Sentinel-1(10 m)and CosmoSkymed(2.5 m)were used to carry out comparative test,the results showed that the detection performance and detection efficiency of the human visual model of ship detection algorithm based on fast Fourier transform was better than that of the algorithm based on discrete Cosine transform,the two-parameter CFAR algorithm and the K distribution CFAR algorithm.