一种基于角点特征的场景锁定新方法

A novel scenelock method based on corner feature

为了消除由飞机旋转、平移等运动造成的场景变化,使机载吊舱能对感兴趣场景连续侦察和跟踪,本文提出一种基于角点特征的场景锁定新方法。本文将参考图像等分成16块,选取每块中响应值最大的角点作为特征点,采用SAD(sum of absolute differences)算法进行特征点匹配,然后利用改进的RANSAC算法进行角点的一致性检测,利用最小二乘法解算全局运动参数,最后计算图像之间的累积运动。该算法便于在DSP+FPGA硬件平台上实现,能在20 ms内完成处理,实验结果表明,该算法匹配精度高,稳定性好,能较好地实现场景锁定。

In order to solve the problem of scene change caused by aircraft motion, and make sure that the aircraft pod can continuously track the scene of interest,a novel sceneloek method based on corner feature is presented in this paper. First, the reference image is divided into 16 blocks, and select the corner points with local response maxima which would be matched by sum of absolute differences （SAD）algorithm. Secondly, an improved RANSAC algorithm is proposed to detect the consistency of the corner points. We use least squares method （LSM）to calculate global motion parameters, and finally obtain the cumulative movement between images. This method could be implemented on DSP ＋ FPGA platform with a processing time in 20 ms. The experiment results show that this method could achieve the goal of scenelock with high accuracy and stability.