论文部分内容阅读
针对微结构大范围无损测试问题,提出了一种低重叠度的三维结构拼接方法。首先基于实验参数将结构特征提取区域限制在测量过程中的重叠区域,以减少误匹配出现的可能性并提高计算效率;在上述区域内,通过SIFT(scale-invariant feature transform)算法进行特征点提取;在特征点匹配阶段,根据测量系统参数进一步提出缩小匹配点对搜索范围的方法以提高特征点匹配可靠性;最后以重叠区域的局部连续性为依据,计算校正矩阵以校正测量过程中环境扰动带来的拼接结构错动。实验中,将本方法与目前商业设备的拼接测量功能进行了对比。实验表明,本文方法不但适用于特征丰富的结构,也适用于相似度高的阵列性结构,可在重叠度为6%时实现有效拼接。
Aiming at the large-scale non-destructive testing of microstructures, a low-overlap 3D splicing method is proposed. First, based on the experimental parameters, the feature extraction region is limited to the overlapping region in the measurement process to reduce the possibility of mismatch and improve the computational efficiency. In the above region, the feature points are extracted by scale-invariant feature transform (SIFT) In the stage of feature point matching, according to the measurement system parameters, the method of reducing the matching point to the search range is further proposed to improve the matching reliability of feature points. Finally, based on the local continuity of overlapping regions, the correction matrix is calculated to correct the environmental disturbance Bring the mosaic structure staggered. In the experiment, this method is compared with the splicing measurement function of the current commercial equipment. Experiments show that the proposed method is not only applicable to feature-rich structures but also to array structures with high similarity, which can be effectively spliced when the degree of overlap is 6%.