摘要
在红外序列图像中识别规则变化小目标时,传统Top-hat算法是常用的目标检测方法,但是可能会检测到很多虚假目标,而且目标发生变化时不能稳定检测。为了准确检测目标,提出一种红外序列图像中规则变化的小目标检测方法。首先对序列图像利用SIFT算法提取特征点,然后利用RANSAC算法进行特征匹配,拼接形成全景图像,结合对单帧图像利用改进的Top-hat算法进行小目标检测,并且标记,最后根据标记的统计结果得到真实目标位置。
Traditional Top-hat algorithm is a common method of target detection for recognising small objects with regular changes in infrared image sequences.But it is ineffective when the shape of objects is changed,and would yield false results.In order to detect the targets accurately,a detection algorithm for small targets with regular changes in infrared image sequence is presented.First the SIFT algorithm is employed to extract the feature points,and the RANSAC algorithm is used to match the features,then a panoramic image could be stitched.Secondly,the modified Top-hat algorithm is used for detecting small targets in combination of single frame image,and the detected ones are marked.At last,the locations of real targets are found out according to the statistic results of these marks.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第2期266-268,共3页
Computer Applications and Software