J Shanghai Jiaotong Univ Sci ›› 2023, Vol. 28 ›› Issue (5): 560-568.doi: 10.1007/s12204-022-2476-8

• • 上一篇    下一篇



  1. (上海交通大学 仪器科学与工程系,上海200240)
  • 接受日期:2020-12-23 出版日期:2023-10-20 发布日期:2023-10-20

High Curvature Stripe Profile Extraction Algorithm of Line Structured Light Measuring System

SUN Hao (孙昊),DU Xuan (杜宣), LÜ Na(吕娜), CUI Bin(崔斌), ZHA0 Hui*(赵辉)   

  1. (Department of Instrument Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China)
  • Accepted:2020-12-23 Online:2023-10-20 Published:2023-10-20

摘要: 在线结构光轮廓测量系统中,轮廓条纹的解析质量直接影响测量精度,尤其对于具有高反射率和高曲率的表面,轮廓条纹提取算法的解析能力显得尤为重要。本文研究了线结构光的成像原理、轮廓条纹的光强分布规律,并提出了一种基于实际光强分布的条纹轮廓提取方法。在算法中,对条纹图像进行了感兴趣区域快速提取、条纹宽度估计以及自适应滤波;在亚像素精度下提取了条纹的能量中心;在滤除低质量中心点后,利用上下文信息恢复了缺失中心点。使用基于线结构光成像原理生成的模拟图像以及真实实验图像评估了此方法的准确性和重复性。结果表明,该方法对高曲率条纹边缘的最大提取误差为1.6像素,是经典Steger算法的10倍;实验重复精度为8.8微米,是Steger方法的2.7倍。本文提出的方法提高了激光轮廓提取的准确性,适用于高曲率物体的激光轮廓检测。

关键词: 轮廓检测, 线结构光, 轮廓提取, 亚像素, 高曲率

Abstract: In the line structured light measuring system, the accuracy of the process of laser stripe directly affects the measurement results. Therefore, the extraction algorithm for the laser stripe, especially the surface with high reflection and high curvature, is very important. The imaging principle of line structured light, the light intensity distribution law of laser stripe and the extraction algorithm have been studied, and a stripe profile extraction method based on real light intensity distribution has been proposed. In this algorithm, fast region of interest extraction, stripe width estimation, and adaptive filtering on the striped image are performed. Then the energy center of the stripe at the sub-pixel level is extracted. Finally, the low-quality center points are eliminated, and the context information is used to recover the missing central points. Simulated images generated based on the imaging principle of line structured light and real experimental images were used to evaluate the accuracy and repeatability of the proposed method. The results show that the method behaves excellently at the edges of high-curvature stripes; the maximum error is only 1.6 pixels, which is 1/10 of the classic Steger algorithm; the experiment repeatability is only 8.8 μm, which is 2.7 times that of the Steger method. Therefore, the proposed method improves the accuracy of object contour extraction, and it is especially suitable for contour detection of objects with high curvature.