融合区域匹配的快速多线结构光测量方法

Fast Multi-Line Structured Light Measurement Method Integrated with Region Matching

针对多线结构光测量中测量效率与测量精度相矛盾的问题,提出了一种融合区域数字图像相关匹配的快速三维测量方法。采用双边滤波和大津阈值法对图像进行预处理,利用重心法提取初始条纹中心点。用双三次插值法提升初始点的局部邻域分辨率,在局部邻域内再次利用灰度重心法提高中心点的提取精度。以左图像提取的中心线为基准,利用数字图像相关法在右图像中心线附近搜索基准对应的立体匹配点,大大缩小了数字图像相关法的搜索范围,提高了测量效率和立体匹配精度。对标准块规的测量实验结果表明,该方法的中心线提取速度为Steger法的3.29倍,均方根误差相比传统极线约束立体匹配法减小了68.57%。

According to the contradiction between measurement efficiency and accuracy in the multi-line structured light measurement, a fast measurement results of the measurement method base on region digital image correlation matching is proposed in this paper. The bilateral filter and Otsu thresh method are used to preprocess the image, and the gray centroid method is utilized to extract the initial point of the stripe center. The bicubic interpolation method is adopted to improve the local neighborhood resolution of the center initial point for the reuse of the gray centroid method in the local neighborhood to improves the accuracy of the center point extraction. Based on the center line extracted from the left image, the digital image correlation method is used to search for stereo matching points corresponding to the reference near the center line of the right image, which greatly reduced the search range of the digital image correlation method and improved measurement efficiency and stereo matching accuracy. The measurement experimental result of the standard block gauge show that the centerline extraction speed of the method is 3. 29 times that of the Steger method, and compared with the traditional epipolar constrained stereo matching method, the root mean square error of this method is reduced by 68. 57%.