Gabor小波目标特征提取和跟踪方法的研究

A target feature extraction and tracking method using Gabor wavelet

为了解决目标在发生旋转、尺度及平移等几何变形时的自动跟踪问题,提出 Sobel 边缘检测与 L-M 优化结合的 Gabor 小波目标提取与跟踪方法。该方法自动寻找目标的小波特征点,在跟踪过程中不断优化特征模板的旋转、尺度以及平移等几何变形参数,使得小波特征向量值在最小平方和意义上与初始值相匹配,实现目标跟踪。实验结果表明,该算法能够有效地提取目标特征,在目标尺度变化达到 50%,角度变化 10o,旋转 90o等不利条件下实现了可靠跟踪。

In order to extract Gabor wavelets target features and realize stable tracking while the target is rotating, scaling and translating. A new Gabor wavelet tracking method that uses L-M optimization and Sobel edge detection is proposed. It finds that wavelet feature points automatically keep optimizing the rotation, scale and translation parameters of feature model during tracking to make the feature vector match its original values in a manner of sum of squared differences. Consequently, automatic tracking of target resistant to changes of target’s orientation, scale and translation is realized. Experiments show that target feature can be extracted effectively with this method. Under the situation of target scale changing 50 percent and angle changing 10 degree and rotating 90 degree, this method still works.