摘要
提出了一种从频率域出发,估计出运动物体在空间域上位移的算法.利用傅里叶变换的自配准性质和位移特性,用极坐标形式下连续图像相位谱的差直接估计运动目标的位移,并进一步推广用于解决高速公路上视频交通监测中车速的非接触式、动态和高精度测量.与传统的寻找迪拉克峰值的方法相比,该算法先求出相位谱的差,再用其周期数来估计位移的算法简单明了,由于其无需重新变换回空间域,从而节省了处理时间,具有更好的实时性,且精度高,实验证明其分辩能力不低于一个像素,其平均测速精度不低于95%.
An approach utilized to estimate the translation of a moving object in spatial domain based on its Fourier transform spectrum is proposed. According to the phase shift character of Fourier transform, the spatial displacement with the phase spectrum difference between two continuous images of a moving object in the polar system is directly measured. The approach is applied for the un-touched, dynamic and high-precision measurement of vehicle velocity on highway. Being compared with the tra- ditional algorithm of Direc Peak Searching, this approach does not need to transform the spectrum into spatial domain, which reduces the processing time spend on velocity measurement. Additionally the accuracy of system is also improved. The experimental results show that the algorithm can provide the vehicle displacement with a resolution less than one pixel and the precision of mean velocity is more than 95 %
出处
《广州大学学报（自然科学版）》
CAS
2007年第5期13-16,共4页
Journal of Guangzhou University:Natural Science Edition
关键词
傅里叶变换
相位谱差
交通流
车速测量
Fourier transform, phase spectrum difference, traffic flow, vehicle velocity measurement
