一种基于SIFT特征光流的运动目标跟踪算法

Algorithm of Moving Target Tracking Based on SIFT Feature Optical Flow

针对传统光流跟踪算法计算复杂度高、受噪声影响大的问题,提出了一种基于尺度不变特征变换(Scale Invariant Feature Transform,SIFT)和卡尔曼滤波器的特征点光流跟踪算法。首先,利用SIFT算法提取图像中的特征点;然后,根据最小绝对值误差准则对运动目标的特征点进行匹配,建立卡尔曼滤波器方程来计算特征点光流;最后,通过光流特征聚类实现运动目标的识别与跟踪。实验结果表明,算法对自然场景中的运动目标具有良好的跟踪特性,稳定性好,计算量小,易于实现。

The available feature-optical-flow algorithms have great shortages of computing complexity and anti-noise performance. Concerning this problem, a moving target tracking algorithm based on scale invariant feature transform and Kalman filter algorithm was proposed. First, the SIFT features are extracted in images. Then, the feature points of moving target are matched according to the minimum absolute error criterion and the optical flow vectors of SIFT fear tures are estimated by Kalman filter algorithm. Finally, recognition and tracking of moving target are achieved using the clustering algorithm based on optical features. The experimental results suggest that the algorithm performs well on the feature points tracking in natural scene. The algorithm is easy to calculate and achieve.