首页 | 本学科首页   官方微博 | 高级检索  
     

一种特征点跟踪的运动目标检测
引用本文:王俊卿,史泽林,黄莎白.一种特征点跟踪的运动目标检测[J].光电工程,2005,32(9):12-15,30.
作者姓名:王俊卿  史泽林  黄莎白
作者单位:1. 中国科学院沈阳自动化研究所,辽宁,沈阳,110016;中国科学院研究生院,北京,100039
2. 中国科学院沈阳自动化研究所,辽宁,沈阳,110016
摘    要:提出一种基于特征点跟踪的运动目标分割算法。在角点跟踪过程中,8等分匹配点邻域,构造方向子邻域提高匹配精度;提出一种新的聚类准则抽取最优特征子集估计运动参数;结合统计方法消除残差噪声;采用时域滤波滤除孤立区域检测出运动目标。对200帧可见光视频序列和100帧红外图像序列中运动汽车进行检测分割实验,检出率分别达到96%和94%。

关 键 词:目标检测  目标分割  运动目标  角点检测
文章编号:1003-501X(2005)09-0012-04
收稿时间:2004-12-20
修稿时间:2004-12-202005-07-15

Detection of a moving object via feature tracking
WANG Jun-qing,SHI Ze-lin,HUANG Sha-bai.Detection of a moving object via feature tracking[J].Opto-Electronic Engineering,2005,32(9):12-15,30.
Authors:WANG Jun-qing  SHI Ze-lin  HUANG Sha-bai
Affiliation:1. Shenyang Institute of Automation, the Chinese Academy of Sciences, Shenyang 110016, China; 2. Graduate School of the Chinese Academy of Sciences, Beijing 100039, China
Abstract:This paper proposes a method for moving object detection and segmentation via feature points tracking.Neighbor of the correspondence is divided into 8 equal parts to form an oriental sub-neighbor in the procedure of corner matching.This strategy improves precision of corners matching.A simple and sound model is assumed when estimating object motion.Then,a new rule of cluster is proposed to extract the best subset of corner pairs which can be used to estimate motion model parameters.Noises of residue image are removed by statistical method to segment moving object.Some island areas are filtered by temporal filter in post-processing stage.Two image sequences including 200 frames of optical images and 100 frames of infrared image are detected and the detection ratio achieves 96% and 94% respectively.
Keywords:Target detection  Target segmentation  Moving object  Corner detection  
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号