With the advantages of high measurement accuracy, fast measurement speed, large measuring range and non-contact skill, the multi-vision measurement system is widely used in aerospace, automotive and other fields of dynamic target space high-precision positioning. However, due to the large amount of image data and the high complexity of matching and reconstruction algorithms, the real-time performance of the system faces the challenge. Therefore, this paper proposes a distributed multi-eye vision measurement System based on the ZYNQ multi-processor system on chip (MPSOC) platform, and optimizes the architecture of algorithms such as image acquisition, marker point matching, and beam method adjustment 3D reconstruction. By means of matrix block processing, constructing task-level pipelines and other methods to reduce computing delay and resource consumption, an efficient system hardware architecture was built and deployed to the ZYNQ MPSOC platform. The experimental results show that the real-time measurement of the spatial position of four or more high-resolution industrial cameras could be up to 42.3 fps at 2 048×2 048×8 bit, and the average reprojection error of the three-dimensional coordinate of the target is better than 0.72 pixels. In the dynamic tracking measurement experiment for the marker point probe, the maximum error of the system in this paper compared with the C-Track optical dynamic tracking measurement system is 129 μm, and the standard deviation is 43 μm, which can meet the high-precision measurement requirements of dynamic targets.