基于单数码相机的三维视觉测量数据拼接方法

3D data registration method for vision measurement based on a single digital camera

针对大型物体三维形貌视觉测量,提出了一种新的视觉测量三维数据拼接方法。以一台高分辨率数码相机作为全局测量设备,在测量过程中固定不动,并以数码相机坐标系作为全局坐标系。视觉传感器流动到不同位置对物体各子区域进行测量,获得局部坐标系下的三维数据。以一平面靶标为中介将视觉传感器在各位置的局部坐标系统一到全局坐标系下,从而将三维数据统一到全局坐标系下,完成三维拼接。该方法无需在被测物上贴标记,因此适用于任何材质的被测物,且消除了拼接累积误差。实验结果表明,该方法相对拼接误差为0.076%。以Venus石膏像为被测对象,视觉传感器对其三个局部区域进行测量,给出了局部测量结果及拼接结果。

A new 313 data registration method for large surface vision measurement is proposed. A digital camera with high resolution is used as the global measurement equipment, which is fixed at a certain position. The digital camera coordinate frame is taken as the global coordinate frame. The vision sensor measures all sub-areas of the measured object in different positions to acquire 3D data of the sub-areas in local coordinate frames. The local coordinate frames of the vision sensor in different positions are unified into the global coordinate frame by a planar target as the intermediary. Then the 3D data are unified into the global coordinate frame and 3D registration is completed. The proposed method need not stick marks on the measured object, so it is suit to all the measured objects regardless of material. And the method eliminates registration accumulated error. The experimental results show that the relative registration error is 0. 076 %. The vision sensor measures 3 sub-areas of a Venus plaster statue, and the 3D data of sub-areas and the registration results are given.