The springback during production causes spatial deformation in pipes, and the generated assembly stress increases the probability of pipe failure. Considering the safety of pipes, it is quite important to measure the geometric error between the design and the actual pipe before installation in order to prevent potential hazards caused by excessive assembly stresses. However, it is difficult to measure the spatially complex pipes caused by the installation space. In this paper, a monocular vision-based method is proposed for automatic, highly accurate and efficient measurement of spatially complex pipes. The mono-camera is designed to acquire the pipe 2D images, from which local coordinate systems are established based on the measured pipe parameters. Then importantly, transformation matrixes are established, including the transformation between the pixel coordinate system and the local coordinate system, as well as between adjacent local coordinate systems. Next, spatial parameters of pipes in the global coordinate system are obtained through coordinate unification, and 3D models of spatial pipes are reconstructed. Finally, the measurement system is developed and eight repeated experiments of two different types of pipes are performed to verify the precision of the proposed method. The results show that the repeatability deviations of length, bending angle, rotation angle and node coordinate are 0.305 mm, 0.033°, 0.263°, and 0.325 mm, respectively. Compared with the measurement results of the three-coordinate measuring machines, the deviations of length and angle are within 1 mm and 0.5°, respectively. A detailed analysis and discussion of measurement errors are conducted. This study provides a simple technique for automatic, highly accurate, online measurement of pipes.