多源多目标跟踪聚类算法

本文提出了一种改进的多源约束聚类算法，以解决多传感器多目标跟踪（Multi-Object Tracking/Estimation, MOTD）问题。MOTD问题对应于在缺乏噪声和目标运动模型等先验信息的情况下，对多个传感器的量测数据进行聚类。针对现有算法对选定传感器量测敏感的问题，本文提出的算法首先根据选定传感器量测数据点的局部密度，对该传感器量测数据进行筛选排序；其次，对排序后的每一个量测数据点，计算和其他传感器量测的高斯核距离，每个传感器返回距离最小的数据点；最后计算在截断距离内的数据点的数量，当大于给定阈值时判定这些数据点为目标产生的量测，簇的中心（个数）即为目标的位置（个数）。实验结果表明，对比现有多源聚类算法，本文提出的算法在传感器目标检测概率较高的场景中聚类精度和聚类速度均有所改善。 

Multi-Source Multi-Target Tracking Clustering Algorithm

This paper proposes an improved multi-source constrained clustering algorithm to solve the problem of multi-sensor multi-target detection (Multi-Object Tracking/Estimation, MOTD). The MOTD problem corresponds to clustering measurements of multiple sensors without prior information such as noise and target motion models. To solve the problem that the existing algorithm is sensitive to the measurement of the selected sensor, firstly, the proposed algorithm sorts and filters the selected sensor’s measurements based on the local density; secondly, for each sorted data point, calculate the Gaussian kernel distance measured by other sensors, and each sensor returns the data point with the smallest distance; finally, calculate the number of data points within the cutoff distance, when the number of data points is greater than a given threshold, it indicates that all the data points are the target-generated measurements and formed a cluster, the center (number) of the cluster is the target position (number). Experimental results show that, compared with the existing multi-source clustering algorithm, the proposed algorithm improves the clustering accuracy and clustering speed in the scene where the sensor target detection rate is high.