Abstract: To address the issues of dust and oil contamination on tunnel surveillance cameras, as well as the low efficiency and high risk of manual cleaning, a high-pressure water jet automatic cleaning system based on a four-degree-of-freedom (4-DOF) robotic arm is proposed for tunnel electromechanical maintenance. The system is mounted on an aerial work platform of a maintenance vehicle, and a depth camera and high-pressure water jet nozzle are integrated at the end-effector, enabling integrated camera recognition, pose measurement, and targeted cleaning. The Zero-DCE++ low-light enhancement algorithm combined with the YOLOv8 detection model was used to identify cameras under weak illumination. Based on depth images and the vector cross-product method, the mirror normal vector was extracted to determine the spatial pose of the camera. A coordinate mapping relationship between the camera and robotic arm was established, and a two-stage control strategy of “coarse positioning + precise pose alignment” was developed by considering the parabolic characteristics of the water jet. Experimental results show that, after image enhancement, the effective camera detection distance under 4–24 lux increases from 3–11 m to 9–12 m. The positioning errors of translational joints J1 and J2 are less than 0.1 mm, and the angular errors of rotational joints J3 and J4 are less than 1°. A single automatic cleaning operation takes approximately 20 s, representing a 66.7% reduction in operation time compared with manual aerial cleaning, and the cleaning efficiency per unit time is approximately three times that of the manual method. Field tests verify that the system has good engineering adaptability.