摘要
为了提升双目视觉系统三维重建的准确性和实时性,提出了一种基于区域分割和匹配的方法。针对实际场景中存在大面积灰度相近区域的现象以及稠密三维重建存在实时性差的问题,采用分水岭算法提取区域轮廓进行三维重建;针对轮廓边缘的误匹配问题,建立区域匹配和边缘点匹配的双重约束条件进行优化匹配;根据平行轴双目立体视觉模型进行三维重建。结果表明:采用轮廓特征进行匹配因其匹配点数大为减少,匹配用时提高了90%;由于采用了双重匹配策略,匹配和重建的准确性得到了保证。
To improve the accuracy and real-time of the 3D reconstruction method, an algorithm based on region segmentation and matching is proposed. In view of the phenomenon that actual scenes always have large regions with similar gray degree, and the problem that dense 3D reconstruction is poor at real-time performance, watershed algorithm is conducted for extracting region contours. In view of the mismatching problem, dual constraint with region match and pixel match is established for matching optimization. 3D reconstruction is carried out according to parallel stereo vision model. It is proved that, the matching time can be raised by 90% because of the reduction of matching pixels by using contours as matching features, and the accuracy of matching and reconstruction can he guaranteed because of the dual matching constraint.
出处
《光学技术》
CAS
CSCD
北大核心
2015年第4期322-326,330,共6页
Optical Technique
基金
"十二五"预研重大项目(40401060305)
关键词
梯度图
分水岭算法
模糊聚类
区域约束
边缘匹配
三维重建
gradient map
watershed algorithm
fuzzy clustering algorithm
region constraint
edge matching
3D reconstruction