摘要
提出了一种用于三维空间中近似平面目标的快速识别方法。该方法利用目标轮廓的曲率信息构造出仿射不变量函数,再利用该不变量的极值点信息定位出轮廓的起始点进而对不变量做相应的循环移位调整,最终通过对调整后的不变量进行相似度比较实现目标识别。由于曲率信息能够有效地描述目标轮廓特征,所以保证了本方法优良的识别性能。同时本方法通过起始点定位的方式克服了传统上高复杂度的循环移位匹配的识别模式,进而实现了快速识别。实验结果显示了所述方法在识别精度和识别效率上都具有明显的优势。
A fast method for the recognition of planar objects in 3D space is proposed.It is based on an affine invariant constructed by the curvature information of a contour.After the location of starting point and re-arrangement of the constructed affine invariant,the final recognition can be implemented by the direct comparison of the similarity of affine invariant.Since the curvature can describe the contour features efficiently,the constructed invariant can provide a high performance in recognition.Moreover,the high recognition efficiency is obtained by the starting-point location technique instead of the circular shift matching which is of high complexity.The experimental results show high recognition accuracy and efficiency of the proposed method,which implies the superiority on the fast recognition applications.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2010年第7期1384-1388,共5页
Systems Engineering and Electronics
基金
国家自然科学基金与微软亚洲研究院联合(60776795)
国家自然科学基金(60902079
60805012)资助课题
关键词
快速目标识别
三维空间
仿射不变量
曲率
起始点定位
fast object recognition
3D space
affine invariant
curvature
starting-point location
