车载鱼眼成像系统的运动矢量检测方法

Motion-vector detection method of vehicle-borne fish-eye imaging system

在复合成像侦测系统中,鱼眼成像系统的超大视场突破了"旋转扫描"和"多镜头拼接"的设计常规,采用固定凝视的方式对整个半球空域的目标进行搜索。然而在车载环境下,车体运动将导致输出的图像序列不稳定,影响观察效果,甚至会引起黄斑成像系统跟踪偏差。根据鱼眼成像系统的工作特点,通过利用熵值理论确定出图像信息最为丰富的区域,探讨了一种改进的适用于车载鱼眼成像系统运动矢量检测方法。通过实验表明,改进算法在精度上与灰度投影算法相仿,而且速度提高了一个数量级,既保证了鱼眼成像系统搜索目标的实时性,同时又提高了系统的稳定性。

In the compound imaging detection system,because of fish-eye imaging system＇s huge field of view,it breaks through the military optoelectronic device designed convention of ＂rotating scanning＂and ＂multi-mirror splice＂,using the fixed stare style to search the targets of the total hemisphere airspace.However,in the condition of vehicle-borne,the car motion will lead to the output image sequence instability,affect the observation effect,even result in the deviation of yellow-spot imaging system.An improved method of motion-vector detection is approached,it is suitable for vehicle-borne fish-eye imaging system,in the light of the working feature of fish-eye imaging system and making use of the entropy theory to determinate the most informative image area.The experiment indicates that the improved algorithm is the same as the original gray projection in accuracy,but the speed raises a order of magnitude,both assurancing the real-time of fish-eye system searching targets and increasing the stability of the system.