摘要
计算机视觉在物体尺寸测量方面已经得到越来越多的应用。在介绍单目视觉测量原理及其局限性的基础上分析双目立体视觉实现测量目的的优势。描述双目测量的的方法流程。此外针对立体匹配,提出一种NCC与Census融合的立体匹配算法。该算法具有NCC和Census二者的优点,在保持对光照不敏感的同时能较好地示边缘和角点的特征。实验结果表明,该算法速度较快且具有较好的稳定性,测量结果在一定范围内达到测量精度要求,具有一定的应用前景。
Computer vision technology in object size measurements has been more and more applications. Introduces the monocular vision measure-ment principle and its limitations, then analyzes the advantages of binocular stereo vision in object size measurement. Describes the pro-cess of binocular method for measuring. Also for the stereo matching, proposes a stereo matching algorithm, which is a fusion of NCC and Census. The algorithm has the advantages of both NCC and the Census, can efficiently represent the edge and corner features while main-taining the light-insensitivity. Experimental results show that the algorithm is fast and stable, and measurements achieve the accuracy re-quirements within a certain range.
基金
中央高校基本科研业务费(No.106112013CDJZR180015)
国家自然科学基金青年基金((No.61502060)
关键词
计算机视觉
双目视觉
立体匹配
三维测量
Computer Vision
Binocular Vision
Stereo Matching
3D Measurement
