雷达自动目标识别中的HRRP特征提取研究

Study of the HRRP feature extraction in radar automatic target recognition

通过分别考虑基于HRRP一阶统计特性的MCC-TMM(模板匹配分类器)和基于HRRP一阶和二阶统计特性的AGC(自适应高斯分类器),提出了两种新的特征提取方法,使得针对不同的分类器采用不同的特征提取方法,从而更好地滤除HRRP上的高斯噪声和更好地保持不同目标HRRP具有不同的幅度分布即HRRP的可分性。基于外场ISAR和微波暗室两种实测数据的识别实验表明,提出的特征提取方法能够明显提高识别性能。

Based on the classifier that will be used for further classification of the targets, e. g. , the one which uses only the first-order statistics of the HRRPs, i. e. , MCC TMM, and the one which uses the first- and second order statistics of the HRRPs, i. e. , AGC, respectively, two new feature extraction methods which realize the above operations in distinct orders are proposed, for filtering the Gaussian noises contaminating the HRRPs more effectively while preserving the amplitude differences of the HRRPs between the separate targets. Experimental results on actual measured ISAR data and anechoic chamber data indicate that the extracted features are indeed effective as the inputs to the follow-up first-order statistics based MCC-TMM classifier and the first and second-order statistics based AGC classifier respectively, both resulting in a much higher classification performance coming from the feature extraction process.