Aiming at the problem that the failures of overvoltage protection device are difficult to identify under different levels of noise, and that failure details and device color information are hard to distinguish, a method for automated recognition of the failures of overvoltage protection device based on infrared imaging and Support Vector Machine(SVM) has been studied. Infrared imaging technology is employed to generate infrared thermal images of the overvoltage protection devices. An infrared thermal image enhancement method based on multi-scale Retinex is employed to improve the images. A fault feature extraction method based on wavelet transform is adopted to obtain the fault features of the overvoltage protection devices. A binary tree algorithm is applied to optimize the SVM, resulting in an SVM binary tree classification model. Automated recognition of the failures of overvoltage protection device is achieved through model training and solving. Experimental results show that the enhanced infrared thermal images of the overvoltage protection devices retain good detail and color information, with a Peak Signal-to-Noise Ratio(PSNR) increased by more than 50 dB. The extracted fault features of the overvoltage protection devices have good separability. The automated recognition match rate of the failures of overvoltage protection device under different noise levels remains above 96%, allowing for accurate identification of the failures.