摘要
本文提出一种可适应平移、旋转、缩放和仿射变换的二维目标识别算法。使用五点不变量描述图像轮廓,得到五点不变量序列;应用灰色相关分析方法找到模板图像中与目标图像最匹配的五点不变量序列,从而正确识别目标。为验证算法的有效性,选取20个二维目标,每个目标获取14幅从不同视角拍摄的图像,共280幅图像进行识别,结果表明所提出算法较其他相似算法更稳定且识别率更高。
An efficient model-based recognition algorithm is proposed to recognize 2D objects under translation,rotation,scale transformation and affine transformation.Five-point invariant descriptor is used here to represent the contours of objects.Grey relational analysis is then employed to classify the objects by analyzing the relationship between them.To verify the proposed method,20 2D objects are used,and 14 images for each object are taken from different angles of camera view.So,there are 280 images to be tested.The experimental results reveal that the proposed method has a higher correct rate of recognition,and is more effective and reliable in object recognition compared with similar algorithms.
出处
《广西师范大学学报(自然科学版)》
CAS
北大核心
2010年第4期30-33,共4页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金重大课题资助项目(90718020)
广西研究生教育创新计划资助项目
关键词
二维目标识别
五点不变量
灰色相关分析
2D object recognition
five-point invariant descriptor
grey relational analysis