摘要
立体视觉匹配一直是机器视觉和模式识别领域中的一个重要问题。极线约束是立体匹配中的基本约束之一,为进一步提高立体匹配精度和效率,在对极线约束数学描述推导的基础上,通过求解基本矩阵得到相应匹配点的极线,提出了求解基本矩阵的一般算法,即采用RANSAC算法。首先得到特征点的初始匹配点对,然后运用7点算法将初始匹配点对划分为内点和外点,最后用所有的内点来重新计算优化基本矩阵,并通过对实际图像的立体视觉匹配实验,以验证该匹配算法的有效性。
Stereo matching is an important problem in the field of machine vision and pattern recognition. Epipolar line constraints are often employed in the stereo matching. The mathematic description of the epi- polar line constraints is deduced in this paper. The corresponding epipolar line of the feature points is ob- tained by solving the fundamental matrix in order to increase stereo matching quality and efficiency. Sever- al algorithms are provided for determining the fundamental matrix. The RANSAC algorithm is adopted to acquire the original matching points, and the 7 Points Algorithm is used to classify the original points to in- ner points and outer points. All the inner points are employed to refine the fundamental matrix at last. The matching algorithm is verified by an actual stereo matching exneriment of a couole of oictures.
出处
《组合机床与自动化加工技术》
北大核心
2013年第11期20-22,共3页
Modular Machine Tool & Automatic Manufacturing Technique
基金
国家自然科学基金项目(51175233)
江苏省科技成果转化专项资金项目(BA2010068)
江苏省科技支撑计划项目(BE2010060)
镇江市工业科技支撑计划项目(GY2012039)
关键词
极线约束
基本矩阵
RANSAC算法
立体匹配
epipolar line constraints
fundamental matrix
RANSAC algorithm
stereo matching
