海面目标运动速度的视觉估计方法

摘要

In order to conveniently and rapidly measure the speed of maritime moving target with constant speed and direction in planar cartesian coordinates,this paper proposed a reliable method on the basis of binocular vision.First,the proposed method obtained both inner as well as external CCD camera parameters through camera calibration based on computer vision theory,and put forward a mathematical model for measurement.Second,the photoelectric imaging equipment could provide tar-get sequential frame pictures,combined with pixel coordinate of characteristic corners on given images,this method computed the relative target distance and azimuth of instant time by the mathematic measuring model.Finally,according to the geometric relation,this paper introduced pseudo linear method to transform motion equation so as to calculate the target velocity with the least square estimation algorithm.This paper accomplished massive numbers of computer simulation to research the effect that different variables might have.Computer simulation results indicate that estimation accuracy of 3000 m target with speed of 10~30 Kn is less than 0.2 m/s.The proposed method has very few restrictions and is easy to realized,it can provide some val-uable assist for covert underwater observer to judge the war field situation for quick decision-making as well as offshore surveil-lance.%为方便快速地实现对平面直角导航坐标系下匀速直航海面目标的速度测量，提出一种立体视觉估计方法。该方法基于计算机视觉理论标定得到摄像机参数（包括内参数及外参数），建立立体视觉测量数学模型，然后利用光电成像设备连续获取目标图像序列，结合图像目标特征角点信息测量被测目标相对距离，同时计算得到不同时刻的目标方位。在此基础上，根据目标运动几何关系，采用拟线性方法处理目标运动方程，最后实现对目标速度的最小二乘估计。通过大量仿真实验，研究了不同初始条件对目标速度估计性能的影响。结果表明，对于3000 m以内航速在10～30 Kn的海面目标，估计精度优于0．2 m／s。该方法要求简单，便于实现，所得结论可为岸基监控海面目标或水下隐蔽观测站仅利用图像信息快速判断被测目标运动态势及作战决策提供参考。