线结构光条纹中心亚像素自适应提取算法

为了解决传统条纹中心提取算法对物体材质及噪声敏感问题, 采用自适应结构光条纹中心提取算法来提取条纹亚像素坐标。该算法首先对图像进行预处理, 利用图像掩模操作提取条纹图像感兴趣区域, 通过自适应卷积模板消除噪声干扰, 得到条纹区域截面宽度及条纹中心坐标的像素集合; 其次根据中心坐标的像素集合采用二次加权灰度重心法求取条纹中心初始坐标值, 将此作为种子点进行区域生长运算, 并结合主成分分析分解特征矩阵; 最终得到线结构光中心的亚像素坐标点。结果表明, 该算法能够有效快速地获取结构光条纹中心亚像素坐标, 相比灰度重心法, 所提算法实验结果波动性较小且标准误差也相对较小, 提取速度相比基于Hessian矩阵的Steger法提高近4倍, 满足工业测量系统实时性要求。所提出的结构光条纹中心提取算法具有较高的提取精度和良好的稳健性, 计算复杂度低, 具有较高的实时性, 可为后续3维视觉测量系统提供良好的精度保障。

Sub-pixel adaptive center extraction of line structured light stripe

In order to solve the problem that the traditional stripe center extraction algorithm is sensitive to material and noise, an adaptive structured light stripe center extraction algorithm was used to extract the fringe sub-pixel coordinates. The algorithm first preprocesses the image, extracts the region of interest of the stripe image by using the image mask operation, eliminates noise interference through the adaptive convolution template, and obtains the pixel sets of the stripe area cross-sectional width and center coordinates. Secondly, according to the pixel set of the central coordinates, the initial coordinate value of the stripe center was calculated by the quadratic weighted gray centroid method, which will be used as the seed point for regional growth operation, then combined with principal component analysis to decompose the characteristic matrix, and finally the sub-pixel coordinate point of the center of the linear structured light was obtained. The results show that the center sub-pixel coordinates of the structured light stripe can be effectively and quickly obtained by this algorithm. Compared with the gray-scale barycenter method, the extraction results of the algorithm in this paper are less volatile and have a relatively small standard error. The extraction speed is nearly 4 times higher than that of the Steger method based on Hessian matrix, which meets the real-time requirements of the industrial measurement system. The proposed algorithm in this paper has high extraction accuracy, good robustness, low computational complexity, and high real-time performance, which provides nice accuracy guarantee for the subsequent 3-D vision measurement system.