在基于高分辨距离像的雷达目标识别中,幂变换是一种常用的非线性预处理方法,对提高识别率有一定的作用。但幂变换对高分辨距离像目标识别的性能提升受信噪比情况影响较大,且幂变换指数的选取对识别结果影响明显,甚至有不利影响。文中针对幂变换存在的问题,提出用对数变换对高分辨距离像进行预处理。比较分析了对数变换与幂变换的作用机理的异同点,并重点分析了对数变换对高分辨距离像类内和类间相关系数的影响。研究表明:对数变换相较于幂变换更适于对高分辨距离像进行预处理,且实测数据的实验结果表明对数变换比幂变换能更有效地提高识别率。

In radar target recognition based on high resolution range profile, power transform is a common nonlinear preprocessing method, which can improve the recognition rate. But the performance of high resolution range profile target recognition with power transform is seriously affected by signal-to-noise ratio, and the recognition result is significantly affected by the choice of the power transform exponential, even the effect is adverse. In this paper, logarithm transform is raised for high resolution range profile preprocessing to solve the problems of power transform, and a comparative analysis of the similarities and differences between the mechanism of logarithm transform and power transform is provided, and the impact of logarithm transform on the intra-class and inter-class correlation coefficients of high resolution range profiles is analyzed. The results show that logarithm transform is more suitable for high resolution range profile preprocessing than power transform, and the experimental results based on measured data show that logarithm transform can more effectively improve the recognition rate than power transform.

TN911.7