计算机科学 ›› 2012, Vol. 39 ›› Issue (Z11): 391-394.

• 图形图像 • 上一篇    下一篇

SIFT算法在CUDA加速下的实时人物识别与定位

罗翌陈,杨辉华,李灵巧,翟雷   

  1. (桂林电子科技大学计算机科学与工程学院 桂林541004) (桂林电子科技大学电子工程与自动化学院 桂林541004)
  • 出版日期:2018-11-16 发布日期:2018-11-16

SIFT Algorithm on the CUDA-accelerated Real-time Character Recognition and Positioning

  • Online:2018-11-16 Published:2018-11-16

摘要: 首先采用Dircctshow技术对视频图像中的运动目标进行差分检测,然后在此基础上对视频图像进行分割和识别,最后实现人体图像及头部图像的识别和定位。视频图像的处理和识别主要采用差分法和SII门特征的模式匹配来完成人体和头部的识别和定位。经实验验证,SIhI’算法在CUDA加速下具有较好的识别速度和准确率,解决了图像识别技术上的一些问题。

关键词: SIFT,模式匹配,图像识别,尺度,特征

Abstract: Firstly, Directshow technology used in the video image to detect moving targets, and on this basis, the video image segmentation and recognition,and finally achieve the image recognition and positioning of the head and the body.Video image processing and recognition using finite difference method and the SIFT characteristics of the pattern mat ching body and head mainly to complete the identification and location. SII门气algorithm on the CUDA-accelerated can get better recognition speed and recognition accuracy by experimental validation, and solve some technical problems of image rccogmtion.

Key words: SIFT,Pattern matching, Image recognition, Scale, Feature

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!