Stereo vision systems and hand-eye systems in an intelligent factory must be able to tolerate vibration in the stereo system and must have high robustness against camera parameter changes during long-term operation. We propose a self-adjustment methodology to achieve maintenance-free stereo cameras. The methodology aims to achieve autonomous recovery of stereo camera rotation parameters, even if the cameras rotate after suffering a weak impact. In autonomous recovery, the self-adjustment procedure modifies the camera rotation parameters to measure objects correctly. We also conduct accuracy verification to validate our approach that stereo camera rotation parameters can be recovered using an active stereo camera in some prepared test environments. Additionally, we verify the validity of our approach through experiments, which include corresponding points acquisition via the correlation coefficient methodology.