Abstract: A method to detect ships with ship-borne radar images is presented in this paper. It adds multiple consecutive radar images, and estimates the probability distribution of the radar backscattering of sea clutter with the Probabilistic Neural Networks(PNN) model. It determines the threshold by applying the Constant False Alarm Rate(CFAR) model and removes the false alarm according to the connected area size of any probable object in the binary image obtained by thresholding. The temporal sequences of X-band ship-borne radar images are used to test the performance of the proposed method. The obtained results show that the detection precision reaches up to 89.5%.

Key words: ship detection, ship-borne radar, Probabilistic Neural Network(PNN), Constant False Alarm Rate(CFAR)

摘要： 提出一种适用于船载雷达图像的船只检测方法，对相邻的多幅雷达图像进行叠加处理，采用概率神经网络模型估计海杂波雷达后向散射的概率分布，利用恒虚警率技术确定全局阈值，根据连通区域的大小去除虚警。使用X波段船载雷达图像序列对该方法进行检验，结果表明，利用该方法得到的船载雷达图像的船只检测精度可达89.5%。

关键词: 船只检测, 船载雷达, 概率神经网络, 恒虚警率

CLC Number: 

• TP18