A method is proposed to identify zero resistance insulators under various pollution levels

humidity andtemperature conditions by combining the insulator infrared image and the support vector machine（SVM）. First of all

the insulator string infrared images got from power system are preprocessed. After the insulator string infrared imagepreprocessing

the two-dimensional Otsu algorithm is used to segment the infrared image

obtaining the binary im-age. The edge detection algorithm is used to extract the edge image of insulator string. The least square method isused to fit the insulator disk edge image

getting the disk elliptic equations. Then the elliptic equation is used to ex-tract the single insulator disk. Temperature features of single insulator disk are extracted. The method of three factoranalysis of variance is used to select features not affected by the insulator polluted degree

environmental humidityand temperature affected parameters as characteristic parameters. Radial basis kernel function is selected as the ker-nel function of support vector machines（SVM）. The Lib VSM tools based on cross validation in the grid search meth-od of parameter selection and training set data are used to determine the kernel function parameter and error price. Fi-nally

the test data are used to test the identification model. The recognition rate obtained is higher

satisfying the re-quirement of the preset.