不完全量测下基于机器视觉的被动跟踪算法

Machine vision based passive tracking algorithm with intermittent observations

针对不完全量测下利用电视摄像机的单站水面目标被动跟踪问题,提出了一种基于机器视觉被动测距的联邦目标跟踪算法.首先,利用机器视觉技术设计了目标距离的被动测量方法;其次,依据测量机理将测量通道分解为机器视觉被动测距通道和传统测角通道,基于验后置信度残差检测的平方根容积卡尔曼滤波(SRCKF)设计了双通道子滤波器,并对子滤波器估计结果进行联邦结构融合得到最终估计结果.通过OpenGL仿真目标图像和真实水面目标视频的测量结果证明了机器视觉测距的有效性,且在不完全量测下,该跟踪算法比传统基于质点的被动跟踪算法具有更高的跟踪精度.

For the problem of passive surface target tracking based on single television camera with intermittent observations,a federal tracking algorithm was proposed by using the distance measurements based on machine vision.Firstly,a machine vision method was developed to obtain the target distance.Then,the detection channel was decomposed into the distance detection channel and bearings detection channel according to the measuring mechanism.Two sub-filters of the detection channels were designed by the posterior confidence residual test based square-root cubature Kalman filter（SRCKF）,and the output of given algorithm was obtained by the federal fusion of two sub-filters′outputs.Finally,the experiment results created by open graphics library（OpenGL）and actual measuring results were demonstrated to show the effectiveness of the provided machine vision solution.And simulation results illustrate that the proposed tracking algorithm provides higher precision than the traditional passive tracking filters for particle target.