摘要
提出了一种以中心矩作为匹配特征,基于积分图像的思想,并采用粗、精阶段相结合的相关跟踪算法。实验结果表明,与以往的以象素作为图像特征的相关跟踪算法不同,本算法运算速度快、实时性强、匹配精度高。采用VisualC++编程,可在40ms,即一个电视周期内完成对图像的处理。本算法在运动目标跟踪、模式识别及机器视觉等领域有着广泛的应用前景。
A new algorithm based on the characteristic of centralized moments and the idea of Integral-Images,which combined two-stage cross correlation tracking algorithms was described. It was different from the traditional algorithms adopting pixel as the characteristic. According to the experiment result,the algorithms could be used to process the image very fast and the matching precision is also high. The matching algorithms was programmed with Visual C(++). It could finish processing the image during the time of 40ms ,which is the period of television. The algorithm will be widdly applied in the future in the field of moving object tracking,pattern recognition and the machine vision.
出处
《计算机应用》
CSCD
北大核心
2004年第11期116-118,共3页
journal of Computer Applications
关键词
中心矩
模板匹配
相关跟踪
积分图像
centralized moments
template matching
correlation tracking
integral-images
