贝叶斯优化与集成学习在弹载雷达目标识别中的应用
Application of Bayesian Optimization and Ensemble Learning in Target Recognition of Missile-Borne Radar
张攀博 1高静 2吴元伟1
作者信息
- 1. 中国空空导弹研究院,河南 洛阳 471009
- 2. 中国空空导弹研究院,河南 洛阳 471009;空基信息感知与融合全国重点实验室,河南 洛阳 471009
- 折叠
摘要
空空导弹在打击低空、超低空目标时,弹载雷达区分目标和杂波的性能下降.本文针对弹载雷达目标识别问题,应用多种集成学习算法和贝叶斯优化算法,建立了多个目标识别模型,测试并对比了模型性能.通过特征提取、数据标准化和特征选择,构建了优选特征的目标杂波数据集.使用贝叶斯优化算法进行调参,构建了XGBoost、LightGBM和CatBoost目标识别模型并测试.测试结果表明,XGBoost、LightGBM、CatBoost的目标识别效果优于随机森林、支持向量机和AdaBoost.选择XG-Boost、LightGBM和CatBoost和随机森林为基分类器,构建了Stacking目标识别模型并测试.测试结果表明,Stacking的目标识别准确率达到98.88%,优于组成它的四个单一模型,但运行效率大幅降低.综合来看,CatBoost的目标识别准确率达到98.03%,虽不是最优,但其测试时间为0.011 s,运行效率的优势更明显.
Abstract
When air-to-air missile strikes low and ultra-low altitude targets,the performance of missile-borne radar to distinguish targets and clutter is reduced.In this paper,aiming at the problem of target recognition of missile-borne radar,multiple target recognition models are established by using a variety of ensemble learning algorithms and Baye-sian optimization algorithm,and their performance is tested and compared.Through feature extraction,data standardi-zation and feature selection,a target clutter data set with optimized features is constructed.The target recognition mod-els of XGBoost,LightGBM and CatBoost are constructed and tested by using Bayesian optimization algorithm to adjus-ting the parameters.The test results show that the target recognition efficiency of XGBoost,LightGBM and CatBoost is better than that of random forest,support vector machine and AdaBoost.XGBoost,LightGBM,CatBoost and random forest are selected as base classifiers,and the target recognition model of Stacking is constructed and tested.The test results show that the target recognition accuracy of the Stacking algorithm is 98.88%,which is better than each of four models that constitute the stacking algorithm,but its operating efficiency is greatly reduced.In summary,the target recognition accuracy of CatBoost algorithm can reach 98.03%.Although it is not optimal,its test time is 0.011 s,and its operation efficiency is more obvious.
关键词
弹载雷达/目标识别/贝叶斯优化/集成学习/XGBoost/LightGBM/CatBoost/StackingKey words
missile-borne radar/target recognition/Bayesian optimization/ensemble learning/XGBoost/Light-GBM/CatBoost/Stacking引用本文复制引用
基金项目
航空科学基金(20200001081006)
出版年
2024
国家科技期刊平台
NSTL
万方数据