为了从扫描图像序列中检测弱小运动目标并对其状态参数进行估计, 提出一种基于随机有限集理论的目标联合检测跟踪算法.根据推扫型光学传感器的扫描特性, 建立目标在像平面的运动模型和测量模型.将目标状态和量测数据描述为随机有限集合, 将目标的联合检测跟踪问题建模为目标状态集的贝叶斯最优估计问题, 并依据随机有限集理论推导出贝叶斯滤波的预测和更新表达式.从算法实现的角度, 利用高斯混合技术实现算法的递推滤波.仿真结果表明, 该算法适应杂波的能力强, 对漏检的影响更小, 可以有效完成推扫型光学传感器的目标检测跟踪任务.

A random finite sets(RFS) theory based joint detection and tracking algorithm was proposed for detecting dim small moving target and estimating its state parameters from scan image sequences. By analyzing the scan characteristics of shave-scan optical sensor, a target dynamic model and observation model were established, respectively. Then target state and measurements was described as a RFS variable. The joint detection and tracking problem was modeled as a Bayesian optimal estimation problem. Prediction and updating formulas of this algorithm were derived using RFS theory. The algorithm implementation problem was taken into account. A Gaussian mixture(GM) implementation is presented. Simulation results show that this algorithm can depress clutters strongly while has small influence on missing detections. It can accomplish the target detection and tracking task efficiently for shave-scan optical sensor.