This paper presents a novel method of intelligent detection of transformer wiring tests. It combines a new deep learning-based object detection algorithm with a tag code identification technique. Complex wiring in the current transformer error test scenarios implies a need for frequent human testing and judgment by digitizing the equipment terminals and the connected wires in the test. The automatic identification of the test connection lines is realized, relying on learning from the standard wiring and logically binding the standard wiring relationship. The proposed method is instrumental in greatly saving labor costs, reducing the possibility of human error, improving work efficiency, and developing a new concept of current transformer error test training for new employees.