首页 | 官方网站   微博 | 高级检索  
     

线段元支撑区主成分相似性约束特征线匹配
引用本文:宋伟东,朱红,王竞雪,张重阳.线段元支撑区主成分相似性约束特征线匹配[J].信号处理,2016,32(8):904-910.
作者姓名:宋伟东  朱红  王竞雪  张重阳
作者单位:1.辽宁工程技术大学测绘与地理科学学院,辽宁阜新 123000
基金项目:辽宁省科技博士启动基金(20141142);国家自然科学基金项目(41101452);辽宁省教育厅科学技术研究一般项目(L2014134)
摘    要:针对特征线匹配中因线段端点不确定而难以提供准确位置的问题,论文提出结合边缘主点与线段元支撑区主成分相似性约束近景影像特征线匹配方法。首先,利用ASIFT算法获取立体像对同名点,计算仿射变换矩阵;其次,对Canny算子边缘检测后的影像进行Freeman链码跟踪,链码分裂获取线段元的边缘主点,将边缘主点视为匹配基元,同时结合仿射变换、核线约束及Harris兴趣值三重约束独立匹配边缘主点;最后,采用线段元支撑区主成分相似性对特征线匹配结果进行一致性检核。论文选取不同类型的近景影像作为实验数据,验证本文算法的有效性和普适性。该算法实现了以点代线匹配,降低直线匹配算法的复杂程度,同时通过线段元支撑区主成分相似性约束检核匹配结果,充分利用近景影像彩色信息,提高了匹配结果的可靠性。 

关 键 词:直线匹配    边缘主点    主成分分析    Freeman链码
收稿时间:2016-01-15

Feature Line Matching Based on the Principal Component Similarity Constraint of Line Segment Primitive Support Region
Abstract:To solve the problem that the accurate location of line segment endpoint is difficult to be identified in line feature matching, a method of combining the main points of the edge and feature line matching based on the principal component similarity constraint of line segment primitive support region is proposed in this paper. Firstly, ASIFT algorithm is used to match the corresponding points and calculate the affine transformation matrix. Secondly, Freeman chain code is applied to obtain the main points of the edge after an edge is detected through the Canny operator, and the main points of the edge are taken as the matching primitive to be matched independently combined with affine transformation, epipolar constraint and Harris interest value constraint. Finally, the initial matching results are checked through the neighborhood principal component similarity consistency of the line segment. The different types of close range images are taken as the experimental data, and the results verify the validity and universality of the algorithm. The method strives to achieve the line matching by points substituting lines, reduce the complexity of the line matching algorithm, use the principal component similarity constraint of the segment line to check the results, and make full use of color information of close images and improve the reliability of the matching results. 
Keywords:
点击此处可从《信号处理》浏览原始摘要信息
点击此处可从《信号处理》下载全文
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-23

京公网安备 11010802026262号