立体视觉曲面重建与系统误差分析

Research on stereo vision for 3D surface reconstruction and system error analysis

采用立体视觉空间曲面重建技术对三维曲面表面成像进行边缘提取、图像匹配、匹配点空间位置计算等步骤,得到三维曲面表面点的空间位置,利用空间点信息对三维曲面形状进行重建,恢复曲面三维形状;并讨论了立体视觉系统的摄像机分辨率、测量范围和摄像机间距等参数之间的关系.利用该方法对堆积物表面形状及体积进行测量实验结果表明,该方法能准确、快速、方便地给出三维曲面的形状.

In this paper, the computer stereo system is discussed for 3D surface reconstrucion. The method utilizes the correlation of images captured by camera from multiple view points to reconstruct 3D surfaces. The reconstruction process mainly includes the following steps： image capture, camera parameters calibration, character extraction, image matching, space point determination, and interpolation. The relationship between camera resolution, metrical area and distance between two cameras is derived and important results are achieved. In lab, the precision of stereo vision is analyzed by accumulation that the volume of accumulation is known. The experiment shows that the stereo vision system can get 3D surfaces more quickly, nicely and conveniently.