During autonomous substation inspections, the inspection camera’s stop position often deviates from its preset location, resulting in misalignment or omission of target devices in the captured images. This study proposes an automatic deskew correction method for camera preset positions, integrating image recognition and fuzzy Proportional-Integral-Derivative (PID) control. The system employs an enhanced Mask R-CNN algorithm to improve image recognition accuracy, facilitating more reliable identification of target objects. Positional deviations between the recognized image and the reference template are computed using a maximum mutual correlation matching technique. These deviations are subsequently used as inputs for an image-based feedback control system. A fuzzy PID control strategy, combined with stepper motor control, is implemented to automatically adjust the camera’s orientation, achieving precise alignment and correction. By addressing the issue of camera misalignment, this approach significantly improves both the accuracy and efficiency of substation inspections, enhancing the reliability of image capture under complex operational conditions.