Abstract: The defects can appear in seal accessories of high-voltage cables not only externally but also internally. It is difficult and inaccurate to conduct full defect detection. Aiming at this problem, a defect detection method of high-voltage cable lead seal accessories based on eddy current detection technology was studied. The eddy current defect detection equipment was placed above the lead seal accessories of high-voltage cables. The eddy current sensor was used to collect electrical signals and filter them to remove the noise in the signals. Two kinds of features in the electrical signals (the marginal spectrum feature and Mel cepstrum coefficient)were extracted to form a feature set. With the feature as input, the probability of each defect was calculated by using the improved random forest algorithm, and the defect type was determined. The results showed that the intersection to union ratio of the proposed method reached its maximum in different types of defect detection, all of which were above 0.8, indicating high detection accuracy.