基于星凸随机超曲面的扩展目标伽马高斯混合势概率假设密度滤波器
Gamma Gaussian-mixture CPHD filter based on star-convex random hypersurface for extended targets

针对杂波和漏检情况下扩展目标形状估计精度低的问题，提出了一种基于星凸随机超曲面（SRHM）的扩展目标伽玛高斯混合势概率假设密度(CPHD)滤波器。该算法在高斯混合概率假设密度滤波的框架下，首先将目标形状建模为星凸随机超曲面，然后通过CPHD滤波估计出目标的质心位置和目标数目，最后通过将目标的质心位置作为目标形状的中心点来结合量测对目标形状进行估计。其中，算法通过自适应估计尺度变换因子和对形状边界进行约束优化，解决了星凸随机超曲面模型存在着边界形状不规则的问题。通过杂波环境下未知数目的扩展目标仿真实验，验证了所提算法的有效性和可行性。
In view of the low accuracy of shape estimation in multiple extended targets tracking in the clutters and missed detections, a Gamma Gaussian-mixture cardinalized probability hypothesis density filter(CPHD) for extended target tracking, which is based on star-convex random hypersurface model（SRHM）, is proposed. Firstly, under the Gaussian-mixture CPHD filter framework, the proposed algorithm models the shape of the target as star-convex random hypersurface model. Then, the CPHD filter is used to estimate the centroid position and the number of targets. Finally, it estimates the shape of the target through the use of measurements by taking the centroid position as the center of the target shape. The algorithm resolves the problem of irregular shape boundary, which exists in the star-convex random hypersurface model by adaptively estimating the scaling factor and using the constrained optimization for shape boundary. The simulation experiment in clutter environment with the unknown number of extended targets validates the effectiveness and feasibility of the proposed algorithm