基于t分布扩展概率主成分分析模型的一维距离像识别方法

发布时间：2018-04-29 07:10

本文选题：雷达目标识别 + 高分辨距离像　；参考：《电子与信息学报》2017年08期

【摘要】：该文针对概率主成分分析(PPCA)模型用于1维高分辨距离像(HRRP)识别对噪声敏感的问题,对经典PPCA模型进行修正。该方法将基于高斯分布的PPCA模型扩展为基于t分布的PPCA模型,能够综合利用t分布对噪声稳健和PPCA模型自由参数少的特性。同时为了减少目标方位敏感性对HRRP统计建模的影响,进一步将t分布模型扩展为混合概率t分布模型,能够以分布趋同的原则将不同方位帧内具有相同统计特性的HRRP数据进行聚类,减少模型的失配,改善识别性能。模型参数通过期望最大值(EM)算法估计,可提高计算效率。最后,通过贝叶斯规则,以获取的统计特征识别测试数据,仿真结果表明该方法能够提高低信噪比条件下PPCA模型的稳健性。
[Abstract]:In order to solve the problem that the probabilistic principal component analysis (PPCA) model is used to recognize the noise sensitivity of the high resolution range profile, the classical PPCA model is modified in this paper. This method extends the PPCA model based on Gao Si distribution to the PPCA model based on t distribution, which can make full use of the robustness of t distribution to noise and the less free parameters of PPCA model. In order to reduce the influence of target azimuth sensitivity on HRRP statistical modeling, the t-distribution model is further extended to the mixed probabilistic t-distribution model. The HRRP data with the same statistical characteristics in different azimuth frames can be clustered according to the principle of distribution convergence, which can reduce the mismatch of the model and improve the recognition performance. The model parameters can be estimated by the expected maximum value (EM) algorithm, and the computational efficiency can be improved. Finally, the Bayesian rules are used to identify the test data. The simulation results show that the proposed method can improve the robustness of the PPCA model under low SNR.
【作者单位】：西北工业大学电子信息学院;
【基金】：国家自然科学基金(61571364) 西北工业大学研究生创意创新种子基金(Z2017022)~~
【分类号】：TN957.52

【参考文献】