基于多传感器不完全量测下的机动目标跟踪算法  

Multi-sensor Information Fusion Motivate Target Tracking Algorithm Based on Missing Measurements

作　　者：李松[1] 胡振涛[1] 李晶[1] 杨昭[1] 金勇[1] 

机构地区：[1]河南大学图像处理与模式识别研究所,开封475001

出　　处：《计算机科学》2013年第8期277-281,共5页Computer Science

基　　金：国家自然科学基金(60972119;61170243);河南省科技人才创新项目(114100510001)资助

摘　　要：针对传感器探测概率小于1的不完全量测情况下的非机动目标跟踪问题,提出一种基于多传感器不完全量测下的扩展Kalman滤波算法。首先,利用残差检测的野值剔除方法,确定目标状态估计过程中传感器是否接收到正确的量测数据;其次,基于每个传感器的量测数据,在不完全量测下采用改进的扩展卡尔曼滤波算法分别对目标运动状态进行估计;进而结合多传感器最优加权融合方法求解基于多传感器观测数据的状态估计;最后,将算法应用到光电跟踪系统中。仿真实验得到不完全量测下传感器探测概率对滤波效果的影响,验证了算法的有效性,其跟踪精度接近完全量测下的状态估计精度。In allusion to the situation with missing measurements in which the sensor detecting probability is less than 1 ,a multi-sensor information fusion motivate target tracking algorithm based on missing measurements was proposed. First of all, based on the algorithm of residual error detection, the wild values from the observed data are distinguished, and accuracy of the measurement data can be determined during the state estimation process of dynamic system. Secondly, according to the improved EKF algorithm based on the conditions with missing measurements, the target motion state is estimated by use of every sensor node＇s measurement data respectively, and then the multi-sensor optimal weighte fu- sion method is utilized to obtain the optimal estimation based on multi-sensor measurement data. Finally, the influence of the probability of the sensor detection on the filtering effect is obtained by simulation experiments using photoelectric sensors. And the simulation results demonstrate the effectiveness of the algorithm. Additionally, the algorithm＇ s trac- king accuracy is mostly approximate to the state estimation accuracy under the situation of complete measurements.

关 键 词：目标跟踪 不完全量测 扩展Kalman滤波 信息融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]