基于角点特征的KLT跟踪全景图像拼接算法

虚拟现实技术在航空航天、机器人遥操作和生物医学等领域得到了广泛应用，已成为研究问题的一种重要手段，全景图像的拼接技术则是其中一个非常重要的环节。通过SUSAN检测算子提取图像特征角点，利用快速RANSAC算法来实现图像的初配准．然后把检测到特征角点作为图像中的运动目标，利用KLT跟踪算法，确定各特征角点的精确位置，从而得到图像之间精确的单应性变换矩阵。在图像融合过程中，提出了消除图像拼接过程中“鬼影”的办法。实验结果表明，该算法能够大大提高图像配准精确度，并具有较好的鲁棒性。

Based on Corner Feature KLT Track Panoramic Mosaic Algorithm

Virtual reality technology is widely used in the fields of aerospace, robot teleoperation and biology medical and so on, and has become an important research way. The panoramic mosaic is an important part of virtual reality, The SUSAN detector was used to extract the feature corner from the images, and fast RANSAC algorithm was used to register the images approximately, Then the detected corners were regarded as moving objects in the images, located accurately by the KLT track algorithm, and accurate homography matrix between the two images was obtained, The method of getting rid of ghost was used in the image mosaic process, The result shows that the accuracy of registration is improved greatly and the method is robust.