基于深度网络的雷达信号分选

Radar Signal Sorting Based on Deep Network

采用传统方法对雷达信号进行分选时,会因为模型准确度低而需要根据外部情况人工调整多个参数以适应电磁环境,而且对复杂环境和多功能雷达的适应性差。基于图像处理方法,将混合雷达信号的脉冲描述字(PDW)编码为图像,选取相应的深度网络,对编码后的图像进行图像语义分割训练,将训练完成的模型用于雷达信号分选,可使信号分选流程智能化,且具有较高的分选准确率。

When the traditional method is used to sort radar signals,it is need to manually adjust several parameters according to the external situation for adapt to the electromagnetic environment because of the low accuracy of the model,and the method has poor adaptability to the complex environment and multifunctional radar.Based on the image processing method,this paper encodes the pulse description words(PDW)of hybrid radar signal into images,selects the corresponding deep network to train the encoded image for image semantic segmentation,and uses the trained model for radar signal sorting,which makes the signal sorting process intelligent and has better sorting accuracy.