Abstract: Based on the characteristic of Fourier transformation, image spectrum takes place related change for rotation, scaling and intensity varying. Aiming at this feature, spectrum information is converted into polar coordinate system. One dimensional invariant based on the statistic of radius is used to ground object target tracking algorithm. The criterion for template updating is defined by target scale transformation based on the statistic of angle. Experimental result proves that the tracking performance of this algorithm is good for image which is rotation, scaling and intensity varying, and its tracking precision is higher and time-consuming is less compared with Nprod and SSAD algorithm.

Key words: Fourier transform, target tracking, spectrum measurement

摘要： 根据傅里叶变换特性，图像在发生尺度、旋转、光照变化时频域内会产生相应变化。针对该特点，将频域信息转换到极坐标系中，通过半径统计度量得到一维不变量用于地物目标跟踪算法，利用角度统计度量判断目标尺度变换，从而确定目标模板更新准则。实验结果表明，该算法对发生尺度、旋转、光照变化的图像有较好的跟踪性能，且相比Nprod, SSAD算法，其跟踪精度较高、耗时较少。

关键词: 傅里叶变换, 目标跟踪, 频谱度量

CLC Number: 

• TP391