介绍了一种能对场景进行稳定锁定的算法,该算法在传统场景锁定算法的基础上进行了改进。当目标进入红外(电视)摄像机视场时,首先对视频图像进行角点检测,提取出具有特征的角点,根据角点间的距离关系,对相邻两场图像的角点进行匹配得到角点对。然后根据角点对间的对应关系解算出相邻两场图像间的全局运动参数,把此参数输入到控制系统,让控制系统对锁定进行闭环,从而实现场景锁定。实验证明,对于传统匹配算法无法稳定跟踪的场景,改进后的算法能够对场景进行稳定跟踪。

An algorithm is introduced in the paper. The stabilized locking of scene can be performed by the algorithm. The algorithm is improved on the basis of the conventional algorithm of scene locking. When the target enters FOV of IR(TV) camera, the corner detection of the video image is performed firstly to extract the characteristics of corner. According to the distance relationship between corners, the corner matching of two neighboring images is carried out to obtain the pair of corner. All the movement parameters between two neighboring images are resolved according to the corresponding relationship between corners. The parameters are input to the control system. The closed loop is performed for locking in the control system to achieve the scene locking. The tests prove that the improved algorithm can realize the stabilization tracking of the scene which the conventional matching algorithm cannot achieve.