As people''s demand for electricity grows more and more, the requirements for corresponding power inspections become higher and higher. A secondary screen cabinet inspection system based on a quadrotor drone is designed. First, in order to solve the indoor GPS rejection problem of the drone, the AprilTag algorithm is used in conjunction with the sensor height information to realize the indoor flight path planning function, attitude correction, and autonomous flight of the drone. Then, the mission program is stored in the onboard Raspberry Pi. This program will use the flight controller and sensors to find and judge the QR code of the mission location, and execute the shooting mission after reaching the designated point. Then, the collected images are transmitted to the ground server through the local area network, and input to the pre-trained CNN network model for anomaly detection. According to the recognition result, it is judged whether the secondary screen cabinet works normally, and the detection result is uploaded to the display platform. The final project was successfully tested in the indoor secondary screen cabinet of Tongren Substation in Guizhou.