改进格雷码结构光双目视觉工件3D测量

3D Measurement of Workpieces Based on an Improved Gray-code Structured Light and Binocular Vision System

摘要:

机器人喷涂、焊接、研磨等现代制造过程,通常需要获取工件3D几何形状信息以实现路径自动规划。针对此需求,提出了一种改进灰度编码的结构光双目视觉成像与三维测量系统。在离线标定获得相机参数的基础上,采用黑白反向重投影方法解决传统格雷码结构光成像中边界模糊问题；同时,提出快速区域块搜索算法,用于立体匹配和视差图生成；最后,采用Lanczos插值算法填补遮挡造成的视差图空白,从而获得完整的深度信息,实现工件高精度3D重建。实际工件成像实验结果表明：该方法能够实现工件快速、高精度三维重建,尤其对无纹理和弱纹理目标能获得较好的结果,能够满足工业环境测量需求。

Abstract:

Modern manufacture processes, such as robot based spray painting, welding and grinding, often need obtain the 3D geometrical and shape information to achieve automatic path planning. To satisfy this requirement, a binocular vision three-dimensional measurement system with an improved gray-code structure light is proposed in this paper. With the camera parameters obtained from an offline calibration, a black-and-white inverse re-projection method is adopted to solve the zy boundary problem of the traditional structured light imaging technique. Simultaneously, a fast region block search strategy is proposed for stereo matching to generate the disparity map. Finally, a Lanczos interpolation algorithm is utilized to fill the vacancy of disparity map usually arising from occlusion. It is shown in the experiments that the proposed method can achieve a quick three-dimensional reconstruction of workpieces, especially for those with non-texture or weak texture, which make it very suitable for measurement under an industrial environment.