摘要
介绍了以时空差分、数据融合和模糊聚类来估计动目标边界点集的技术。为了解决检测所得多目标边界的非闭合问题,提出了一种双向正交多级分割技术。利用空间最小距离准则,实现了各目标边界点集的划分。PC试验说明,在满足最小距离的条件下,该技术能够迅速有效地提取出视频序列中的运动对象。
This paper introduces the segmentation of motive multiple targets in video sequences.Based on spatio-temporal joint, data fuse and fuzzy technology,the method to estimate edges of moving objects in video sequences is introduced. To solve the problem of non-closed edges,the bi-direction orthogonal multi-stage image segmentation technology is developed.Using the minimum-distance criterion,the segmentation of motive multiple targets is fulfilled.PC experiment shows that the technology is able to extract video objects swiftly and efficiently.
出处
《信息与电子工程》
2004年第2期91-94,113,共5页
information and electronic engineering
基金
国家自然科学基金(10376005)
中国工程物理研究院电子工程研究所基金(WF2002-200)
关键词
信息处理技术
多级分割
双向正交
多运动目标
最小距离准则
information processing techniques
multi-stage segmentation
bi-directional orthogonality
motive multiple targets
minimum-distance criterion
