Object tracking in the complex environment based on SIFT
Yu Chengbo, Liu Yu-xuan, Zhang Jing, Ting Yu · 2011
To effective solve the problem about the moving object tracking in the complex environment, we present a method that utilizing the SIFT feature matching algorithm which have the superiorities properties of the scale-invariant, rotation-invariant, using the SIFT algorithm detect the keypoints. To reduce the running time of the SIFT algorithm, we set a threshold, If the ratio of principal curvatures of each candidate keypoints is below the threshold, the keypoints will be kept. Then we though non-negative phase coordinates of the keypoints subtraction to determine the change area, effectively improve the accuracy of the matching. Add the weight to control the model update of the targets, eliminate the errors feature points of matching. Experiments show that this algorithm is effective in improving the performance of objects tracking in the light changed, scale changed or objects rotated. The algorithm has strong robustness.