| 摘要: |
| 针对目前雷达多目标跟踪的评估指标单一化、只能对特定类型关联算法进行评估和单一权重计算导致评估偏差等问题，本文提出了基于宏观和精细的多维度指标体系并结合主客观组合权重，构建雷达多目标跟踪评估模型。首先，分别从多目标跟踪的航迹整体维度与航迹中点迹的精细变化角度入手，提出宏观与精细两类指标对目标跟踪结果进行多维度的分析；然后，针对不同的数据关联方式采用软硬两种决策来对目标与实际航迹进行匹配并计算各指标值；最后，分别通过G1序列法和CRITIC法计算指标的主客观权重，再通过最小二乘法将主客观权重融合为组合权重，将其代入优化模型得到最终的多目标跟踪评估结果。实验结果表明，该模型可以在稀疏、密集等多目标复杂运动场景下对多目标跟踪做出灵活全面的评估。 |
| 关键词: 宏观与精细指标 组合权重 数据关联 软硬决策 |
| DOI： |
| 分类号:TN953 |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目）,中国博士后科学基金 |
|
| Radar multiple-object-tracking performance evaluation model based on multiple-dimensional system and combined weights |
|
|
|
|
| Abstract: |
| Aiming to address issues such as the reliance on a single evaluation index for radar multi-target tracking and the bias caused by using only one specific type of association algorithm and weight calculation, this paper proposes a comprehensive multi-target radar tracking evaluation model. To address these issues, this paper proposes a comprehensive multiple-object-tracking evaluation model based on a macro and fine multi-dimensional index system, incorporating both subjective and objective weights. Firstly, two types of indexes are introduced to analyze the target tracking results from the perspectives of overall track dimension and fine changes in point tracks. Secondly, different data association methods are employed to match targets with actual tracks and cal-culate each index value accordingly. Finally, to ensure comprehensive evaluation results, subjective weights and objective weights for indicators are determined using G1 sequence method and CRITIC method respectively. These weights are then fused into combined weights through least square method before being applied in an optimization model to obtain final multiple-object-tracking evaluation results. Experimental findings demon-strate that this model can flexibly and comprehensively evaluate multiple-object-tracking performance across various scenarios including sparse or dense environments. |
| Key words: macro and fine indicators combined weight data association soft and hard decision |