基于深度学习的海杂波谱参数预测与影响因素分析

该文基于不同雷达参数和海洋环境参数条件下的岸基雷达海杂波实测数据，利用深度神经网络(DNN)建模技术，建立了从多个测量条件参数出发的海杂波多普勒谱参数预测模型，实现了独立于杂波数据、基于环境特征的海杂波谱特征认知，谱频移和展宽的预测精度达90%以上。基于该预测模型，该文提出了一种基于参数循环递减认知的多普勒谱影响因素分析方法，分析了不同测量参数对海杂波多普勒谱预测的影响，得到了谱参数随主要影响因素的变化规律，结果对基于多普勒特征的海面目标检测应用具有重要意义。 

Sea Clutter Spectral Parameters Prediction and Influence Factor Analysis Based on Deep Learning

Using Deep Neural Network (DNN) modeling technology, a prediction model of Doppler spectral parameters of sea clutter based on multiple measurement conditions is established based on measured data of sea clutter from shore-based radar under different radar parameters and marine environmental parameters. The recognition of sea clutter spectral characteristics based on environmental characteristics and independent of clutter data is realized. The spectral frequency shift and broadening prediction accuracy are greater than 90%. Based on the prediction model, an analysis method of Doppler spectrum influence factors based on the parameter cycle decreasing cognition is proposed. The influence of different measurement parameters on the Doppler spectrum prediction of sea clutter is analyzed, and the change law of spectrum parameters with the main influence factors is obtained. The results are of great significance to the application of sea surface target detection based on Doppler characteristics.