边缘主点作用下的立体影像特征线匹配方法研究

Research on the Line Feature Matching in Stereo Image Based on the Main Points of the Edge

提出一种边缘主点作用下的特征线匹配方法,将特征线匹配转化为其线元素上边缘主点的匹配,从而力求实现以点代线、以直代曲的边缘特征线匹配。首先,针对原始影像采用自适应平滑滤波和Wallis滤波进行影像去噪和增强的预处理工作,突出影像现有特征,提高特征提取的准确率,为实现高效可靠的影像匹配奠定基础。然后,采用了一种基于链码准则的特征线提取算法进行特征线及边缘主点的提取。该算法依据八方向链码准则,依次进行边缘检测、链码跟踪、链码分裂,实现物体边缘线的提取,在链码分裂过程中根据距离判断同时获取边缘主点。最后,本文提出一种边缘主点作用下的特征线匹配方法,在目标影像完成边缘主点的提取后,以边缘主点作为匹配基元依次采用单应矩阵约束、核线约束和灰度相似性约束逐步缩小搜索范围,最终确定匹配点,最后成对连接成功匹配的边缘主点还原特征线,进行目视判读剔除错误匹配,最终完成边缘主点作用下的特征线匹配。实验结果表明,本文提出的匹配方法通过相对容易实现的特征点匹配来替代特征线匹配,匹配效果良好,定位精度准确,还可实现不规则边缘特征的匹配。

A characteristic line matching method under the action of the main points of edge is proposed, which transformed the line matching to the main points matching. This method strives to achieve the line matching with the ＂points substitute lines＂ and ＂straight lines substitute curves＂. Firstly, for the noises and low contrast areas in the stereo image, this dissertation adopts the adaptive smoot- hing filter to reduce the image noise and the Wallis filter to enhance the image features. This method highlighting the existing features and improve the accuracy of feature extraction in order to realize the efficient and reliable foundation for image matching. Secondly, this dissertation adopts a line extraction algorithm based on chain code standards. This algorithm achieved the object edge extraction by the edge detection, the chain code tracking and the chain code division based on eight direction chain code standards. The proposed algorithm can not only extract the straight edge of the objects, but also can use straight segment to approximate the boundary of the ir- regular objects efficiently and accurately. Finally, a characteristic line matching method under the action of the main points of edge is proposed. Adopt the homographic matrix constraint, the popular ray constraint and the gray - scale similarity constraint to search the same points of the main points of edge. Finally, connect the main points of edge which was matched successfully; eliminate the false marches by the visual interpretation to achieve line matching. The experimental results show that the proposed method by relatively easy to implement matching feature points instead of characteristic line matching, the matching result is good, accurate positioning pre- cision, also can realize irregular edge feature matching.