A method based on color feature and least squares support vector machine（SVM） model for visible light images of insulators is proposed. By establishing the correspondence relationship between contamination grade and color characteristic

the identification of contamination grade of insulators is realized. Taking the red and brown porcelain insulators in the substation of Nanchang Power Grid as the research target

the color image of insulator is divided by the Otsu to obtain the disk area

and a total of 36 kinds of characteristics of R

G

B

H

S and I component images are calculated. According to Fisher’s discriminant method

the characteristics Save and Smid

which can significantly represent the level of primate filthiness

are screened. The LS-SVM classifier is designed

and the key parameters of LS-SVM are optimized by BAS algorithm. In the end

the discrimination of insulator contamination grade is achieved. The experimental results show that the accuracy of the method determines the contamination grade of insulators is 96. 92 %.