摘要
提出了一种基于尺度不变特征变换（SIFT）算法的大变形测量方法。该方法首先精确地提取出变形前后两幅图像的SIFT特征点，然后生成每个SIFT点的特征描述子，并利用该描述子对SIFT点进行匹配，得到这两幅图像特征点对应的坐标，最后使用仿射变换求出全场位移和应变参数。将该方法运用到模拟散斑图像中，并与基于差分进化算法的DIC方法进行对比。实验结果表明，所提算法不仅能够克服传统方法在转角大于7°容易失效的问题，还可以以较高的精度和效率计算出形变参数。
A method based on scale invariant feature transform（SIFT） algorithm is proposed. First, the SIFT keypoints of two im-ages before and after deformation are extracted precisely. Then, the affine invariant SIFT feature descriptors are constructed and the SIFT keypoints are matched by using the SIFT descriptors, and the corresponding coordinates of the two images are ob-tained. Finally, calculating the strain parameters by using the least squares iterative. It is applied to simulated speckle images and compared with DIC method based on differential evolution algorithm. The experimental results show that the proposed method not only overcomes the disadvantages of the traditional methods can not be used in the surface of the measured object has a rota-tion angle larger than 7° but also can calculate the strain parameters of large rigid body deformation with higher speed and preci-sion.
作者
杨俊林
彭卫
曾小强
YANG Jun-lin,PENG Wei,ZENG Xiao-qiang （School of Business, Sichuan Agricultural University, Chengdu 611830, China）
出处
《电脑知识与技术》
2013年第11期7091-7095,共5页
Computer Knowledge and Technology
基金
四川省教育厅青年基金项目（11ZB058）
