科技通报
科技通報
과기통보
BULLETIN OF SCIENCE AND TECHNOLOGY
2014年
6期
215-217
,共3页
视频图像%阵列%人员定位%像素
視頻圖像%陣列%人員定位%像素
시빈도상%진렬%인원정위%상소
video image array%personnel positioning%pixel
从视频像素灰度序列中提取特定人员的状态特征,采用平面阵对特征信息进行融合处理是实现计算机视觉下特定人员定位的重要步骤。传统的人员视觉定位方法采用视频步进跟踪学习方法,在随机状态分布下的复杂环境人员进行准确定位效果不佳,提出一种基于平面圆阵定位算法的特定人员计算机视觉定位方法,提取监控视频的图像,进行视频图像的背景建模,得到高斯差分尺度空间中的背景像素灰度值,实现对特定人员的表情识别的动作判别,以此为数据基础设计平面圆阵人员定位系统,实现对暴恐等特定人员的准确锁定和跟踪识别。仿真实验表明,采用该算法准确实现1500 m远距离暴恐人员特征识别与定位,误差较小,分辨率和准确性较高。
從視頻像素灰度序列中提取特定人員的狀態特徵,採用平麵陣對特徵信息進行融閤處理是實現計算機視覺下特定人員定位的重要步驟。傳統的人員視覺定位方法採用視頻步進跟蹤學習方法,在隨機狀態分佈下的複雜環境人員進行準確定位效果不佳,提齣一種基于平麵圓陣定位算法的特定人員計算機視覺定位方法,提取鑑控視頻的圖像,進行視頻圖像的揹景建模,得到高斯差分呎度空間中的揹景像素灰度值,實現對特定人員的錶情識彆的動作判彆,以此為數據基礎設計平麵圓陣人員定位繫統,實現對暴恐等特定人員的準確鎖定和跟蹤識彆。倣真實驗錶明,採用該算法準確實現1500 m遠距離暴恐人員特徵識彆與定位,誤差較小,分辨率和準確性較高。
종시빈상소회도서렬중제취특정인원적상태특정,채용평면진대특정신식진행융합처리시실현계산궤시각하특정인원정위적중요보취。전통적인원시각정위방법채용시빈보진근종학습방법,재수궤상태분포하적복잡배경인원진행준학정위효과불가,제출일충기우평면원진정위산법적특정인원계산궤시각정위방법,제취감공시빈적도상,진행시빈도상적배경건모,득도고사차분척도공간중적배경상소회도치,실현대특정인원적표정식별적동작판별,이차위수거기출설계평면원진인원정위계통,실현대폭공등특정인원적준학쇄정화근종식별。방진실험표명,채용해산법준학실현1500 m원거리폭공인원특정식별여정위,오차교소,분변솔화준학성교고。
The extraction of characteristics of specific personnel from video pixel sequence, planar array is used for feature information fusion, this is an important step for the realization of specific personnel positioning. Traditional video step track-ing method is used for positioning, but the positioning effect is bad when the environment is complicated. An improved spe-cific personnel positioning algorithm is proposed based on plane circular array processing, and the video image is extracted, the background modeling of video image is taken, and background pixel gray value in Gauss differential scale space, the fa-cial expression recognition of specific personnel action discrimination are realized. Based on this, plane circular array per-sonnel positioning system is designed, specific personnel accurate locking and tracking identification are realized. Simula-tion results show that this algorithm can achieve long distance violent terrorist personnel feature recognition and positioning as far as 1500 m. The error is lower, and it has high resolution and accuracy.