Transformer secondary circuit test is an important step in the field test of gateway electrical energy metering device. However, the method of finding secondary circuit terminal to be tested is rather complicated in the past. In order to optimize the test process, this paper proposes a terminal number detection method based on EAST. In this method, the training dataset is first established and the EAST model is trained. The trained model is used to detect the text in terminal block image and outputs the size and position information of terminal number region. Then, the region coordinates were clustered by DBSCAN clustering to distinguish possible multi-column terminals, and the slope angle of each column terminal was calculated by linear regression. Finally, combining the slope angle with the average width and height of the region, the corrected detection results of the terminal number region are obtained. Examples show that this method can accurately detect the terminal number in image and effectively improve the efficiency of terminal detection, which lays a foundation for the subsequent secondary circuit’s test.