Object tracking is a technology that is used to calculate object location in the video sequences by tracking the features of the object. LK optical flow is a kind of sparse optical flow which can track object through a few feature keypoints. LK optical flow which is faster than the dense optical flow is widely used in object tracking system.LK optical flow has a better performance by using the image which has rich texture, while the effect is not ideal when the tracking image is in poor quality. This paper introduces an improved tracking algorithm based on LK optical flow. We improved the accuracy of the algorithm by using bilinear interpolation method which can improve the calculation accuracy to sub-pixel level. We have decreased the time spent on detecting keypoint by using AGAST algorithm. We can calucate gradient by using Gauss filter to decrease the Reprojectiton error. The Reprojectiton error can also be reduced by using homography in the process of iteration. By using the method above, the improved algorithm is faster than the original algorithm. Finally the LK SIFT algorithm and the improved optical flow method are combined together to achieve a simple augmented reality system, which verifies the feasibility of the algorithm in augmented reality system. |