摘要
在摄像机运动的情况下 ,提出了一种基于光流场分割和Canny边缘算子融合技术的运动目标检测方法 .这种方法可分为三步 :第一步利用运动的内极线约束和C 均值聚类算法完成目标区域的分割 ,并获得分割图 ;第二步在分割图中利用Canny边缘算子获得细化的目标区域边缘图 ;第三步根据光流场中的流速值完成分割图和边缘图的融合 ,并检测出完整的运动目标 .实验表明 ,这种方法可以有效地从复杂自然场景的图像序列中检测出完整的运动目标 .
A method on detection of moving object using a fusion algorithm based on segmentation of optical flow field and edge extracted by Canny's operator in the image sequence acquired by a moving camera was proposed. This method can be preformed with three steps. In the first step, the segmentation of object area was finished and segmentation map was obtained using motion epipolar constraint and C-Mean value cluster algorithm. In the second step, edge map of object area was established by Canny's edge operator. In the third step, a whole object was detected by a novel fusion method based on segmentation map and edge map. The experiments showed encouraging results using two sequences of natural images with complex backgrounds.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2003年第9期1299-1302,共4页
Acta Electronica Sinica
基金
航天创新基金资助项目 (No.2 0 0 1 )
