无线电工程

针对探地雷达(Ground Penetrating Radar, GPR)因地质或环境因素导致图像出现高密度噪声，有效信号被淹没，进而出现雷达数据读取困难甚至无法识别的问题，提出一种基于Perona-Malik(PM)算法和噪声识别模型(Noise Re-cognition Module, NRM)的降噪方法——PM-NRM。该算法根据GPR数据特点利用改进PM算法加大图像中噪声与背景的数值差异性，使用基于偏微分方程的NRM识别噪点，利用改进中值定理依次对数据中噪点进行恢复，结合迭代条件算法(Iterated Conditional Mode, ICM),以达到降噪的效果。面对高密度噪声数据，同传统的GPR数据降噪方法相比，所提算法在信号单波道拟合度、峰值信噪比(Peak Signal to Noise Ratio, PSNR)和结构相似度(Structural Similarity, SSIM)等客观评价标准中表现较优。实验结果表明，该算法在GPR系统探测工作中具有一定实用价值。

A noise reduction method based on Perona-Malik(PM) algorithm and Noise Recognition Model(NRM)(PM-NRM) is proposed to solve the problems of high-density noise and effective signal being submerged and then appearing difficulty or even failure in the recognition of Ground Penetrating Radar(GPR) data because of geological or environmental influencing factors. According to the characteristics of GPR data, the algorithm uses improved PM model to increase numerical difference between noise and background in the image. Then the algorithm uses the NRM based on partial differential equation to identify noise, and then uses the improved mean value theorem to restore the noise in the data in turn. Finally, Iterated Conditional Mode(ICM) is combined to achieve the effect of noise reduction. For high-density noise data, compared with traditional noise reduction methods of GPR data, the proposed algorithm has better performance in objective evaluation criteria such as single-channel fitting, Peak Signal to Noise Ratio(PSNR) and Structural Similarity(SSIM). Experimental results show that the algorithm has some practical value in the detection work of GPR system.