The detection and type identification of defects in the external insulation of electrical equipment is an important part of high-voltage insulation condition assessment. UV pulse detection technology is used to detect and analyze the discharges by using UV sensors. In this paper, we propose a method to identify the defects of polluted insulator discharge based on the entropy feature of wavelet packet transform. The wavelet packet transform is used to decompose and reconstruct the UV pulse signal of the polluted insulator discharge and extract the three-dimensional time-frequency distribution of the signal for time-frequency analysis, extract the wavelet packet entropy features of the signal in the time-frequency domain, and classify the two discharge types of corona and arc using classifiers such as support vector machine (SVM). In order to verify the effectiveness of the algorithm proposed in this paper, a surface discharge test device was set up in the laboratory, and experiments were conducted to classify the corona and arc discharge types of ceramic insulators, which were detected by UV pulse sensors. The results show that the recognition rate of identifying the two discharge states of corona and arc is as high as 95% using the feature extraction method proposed in this paper. Meanwhile, the detection system can realize long-term online detection of external insulation discharge and can be extended to substation and switchgear detection.