基于谱聚类的自适应新生目标强度状态提取

Method of adaptive newborn intensity state extraction based on spectral clustering

由于K均值算法易受噪声干扰且对初始粒子的选取较为敏感,在进行PHD状态提取时,难以获得稳定可靠的状态估计结果。对此,提出基于谱聚类的自适应新生目标强度状态提取方法,运用核密度估计理论和mean-shift算法二次估计PHD,提取滤波分布的峰值位置作为各个目标状态的估计值进行目标状态的提取。实验结果表明,改进后的算法相比原始算法在精度上有了明显提高。

The clustering algorithm has great dependence on the initial chosen particles and may be disturbed by noise around the circumstance, thus it affects the results of the state extracting. To improve the accuracy of the state extraction and reduce the burden of the calculation, a brief approach was adopted to extract the state of the targets, based on the adaptive newborn target intensity PHD filter. To further increase the filter precision, the spectral clustering algorithm was designed to estimate PHD with kernel density estimation theory and mean shift method. The peak value of the PHD filter was used as the estimation value to get the state of the targets. The simulation results show that the improved algorithm has better performance than the original algorithm.