应用角点匹配实现目标跟踪

Target tracking using corner matching

角点特征是图像的一个重要局部特征,因其具有计算量小、匹配简单以及旋转、平移、缩放不变的性质,而在图像配准与匹配、目标识别、运动分析、目标跟踪等应用领域都起着非常重要的作用。本文提出了一种新的基于角点特征的向量匹配方法,该方法利用Harris算子检测出目标角点,通过角点的矩特征形成目标的特征向量,最后通过对序列图像的目标特征向量进行匹配来实现目标跟踪。此算法在一般情况下能匹配80%以上角点,在遮挡情况下仍能正确匹配70%左右,处理速度达到20frames/s,满足了实时要求。实验结果证明此方法可有效地抵御目标的变形和遮挡情况。

The comer feature is the important feature of an image. The corner feature needs small calculated amount and can be matched easily. Moreover, the comer feature can keep the invariance after being rotated, translated or scaled. Therefore, it plays an important role in the application fields such as image registration and matching, object recognition, motion analysis and target tracking. A new method of vector matching based on the comer feature is presented in this paper. Firstly, a Harris operator is used to extract the corners of a target, and then the feature vector of the target comes out from the moment feature of comers. Finally, the match feature vector in the sequence images can be used to track the target. Using this method, more than 80% of comers in the normal condition and 70% of comers in the shelter conditions can be matched. Processing speed can achieve 20 frames/s to meet the real-time requirements. Experiment result shows that this method can effectively resist the deformation of the target, shelter and so on.