In order to study the seismic damage detection method of Ultra-high voltage (UHV) transformer bushings, this paper proposes to use high-pass filtering on the acceleration response of the structure and then obtain the filtered scalogram through continuous wavelet transform, which are used as the characteristic parameter of damage. The acceleration responses of the bushing under different damage conditions are calculated by finite element method(FEM), which are input as training data into the convolutional neural network, and the accuracy of this method is verified by the results of shaking table tests. The results show that this method has excellent noise resistance and accurate identification. Abnormal high-frequency information of acceleration response caused by seismic structural damage can be used as damage characteristic.The information of single damage on bushing structure is contained in signals from different positions.This method need not to compare with the undamaged structural response, which means it can rapidly identify the structure damage of UHV bushing caused by earthquake.