【摘要】 高分辨距离像（HRRP）反映了目标空间散射结构在雷达视线方向的投影，近年来被认为是地面目标识别的重要途径。现有的HRRP识别方法采用手工特征加传统机器学习分类器，均属于平面分类方法，即采用统一标准不加区别的优选特征并单次决策最终类别。然而该方法在实际应用中面临种类繁杂、数据不平衡、HRRP姿态敏感性等诸多问题，难以获取最佳的应用效果。层次化方法采取分而治之思想，将一个复杂的细粒度识别任务拆解为多个简单的识别子任务。本文采用层次化识别的思路，提出了一种基于语义引导层次化分类的雷达地面目标识别方法。该方法以联合语义和数据构建的树形结构将一个复杂的细粒度识别任务拆解为多个简单的识别子任务，并针对每一个识别子任务匹配一套优选特征集和一个局部分类器。本方法在仿真数据和实测数据上完成了验证。实验结果表明了本文方法处理地面目标识别任务的有效性。更多还原

【Abstract】 High-Resolution Range Profile(HRRP) is increasingly recognized as a critical method for ground target identification, reflecting the spatial scattering structure of targets along the radar line of sight. Traditional HRRP identification techniques typically employ hand-crafted features and conventional machine learning classifiers in a flat classification approach, applying a uniform set of preferred features and making a single decision on the final category. However, this approach faces significant challenges in practical applications due to complex target categories, data imbalance, and sensitivity to HRRP postures, often resulting in suboptimal performance. To address these issues, this paper introduces a novel method for radar ground target identification based on a semantically guided hierarchical classification approach.This method adopts a divide-and-conquer strategy, effectively breaking down a complex, fine-grained identification task into multiple, more manageable sub-tasks. It employs a tree structure, jointly constructed using semantic and datadriven information. Each sub-task is matched with a tailored set of optimal features and a local classifier, ensuring a more nuanced and effective approach to target identification. The proposed method has been thoroughly tested and validated using both simulated and real-world data. The experimental results demonstrate the efficacy of this approach in handling ground target identification tasks, significantly enhancing accuracy and robustness compared to traditional methods. This semantically-informed hierarchical approach opens new avenues for advanced ground target identification, providing a robust framework for tackling the inherent complexities in HRRP data.更多还原