基于参数化统计模型的雷达HRRP目标识别方法综述

现代战争日趋信息化和智能化，雷达自动目标识别技术(RATR)在国家安全防卫和战略预警等军事应用方面发挥着更加重要的作用。高分辨距离像(HRRP)反映了目标散射点沿雷达视线方向的分布情况，包含了目标丰富的结构信息，对目标识别十分有价值，已成为RATR领域的研究热点。参数化统计建模旨在构建参数化数学模型表征观测数据的分布特性，是估计数据概率分布和挖掘数据隐含信息的重要手段。基于参数化统计模型的雷达HRRP目标识别就是在对HRRP参数化统计建模的基础上，直接利用估计的概率分布进行统计识别或将获取的隐含信息输入分类器进行识别。由于模型具有可融入一定的先验知识、扩展灵活、提供待求参数的不确定性评价以及能结合贝叶斯理论实现自动定阶等优势，基于参数化统计模型的HRRP识别方法整体识别性能优于其他方法，是目前HRRP识别的重点研究方向。该文从浅层和深层参数化统计建模两方面，对近15年的雷达HRRP目标识别方法进行了归纳总结，并分析了各类方法的特点和存在的问题，最后对基于HRRP参数化统计建模的雷达目标识别发展方向进行了展望。 

Survey of Radar HRRP Target Recognition Based on Parametric Statistical Model

In the gradually becoming information-based and intelligent modern warfare, Radar Automatic Target Recognition (RATR) technology plays an increasingly important role in military applications, such as national security defense and strategic early warning. The High-Resolution Range Profile (HRRP) reflects the distribution of target scatterers along the radar line of sight and contains a target’s rich structural information, thus being valuable for target recognition and having become a research hotspot in the field of RATR. Parametric statistical modeling aims to construct a parametric mathematical model to characterize the distribution of observed data. It is an important way to estimate the data probability distribution and mine the hidden information of data. Radar HRRP target recognition based on a parametric statistical model directly uses the estimated probability distribution for statistical recognition or inputs the extracted information hidden in data into the classifier for target recognition. The parametric statistical model exhibits advantages in prior knowledge integration, flexible expansion, parameter uncertainty evaluation, and automatic order determination combined with Bayesian theory; therefore, the overall performance of the HRRP recognition method based on such a model is better than that of other methods. Therefore, parametric statistical modeling is currently the key research direction for radar HRRP recognition. This paper summarizes the radar HRRP target recognition methods of the last 15 years from the two aspects of shallow statistical modeling and deep statistical modeling, analyzes the characteristics and problems of these methods, and forecasts the development direction of radar target recognition based on HRRP parametric statistical modeling.