The invention discloses an automatic identification method for detecting the hydrophobic grade of a composite insulator by using image features. According to the automatic identification method of theinvention, pigments are added, so that contrast can be improved; water drops/water films which are extracted from the colors of stains and the insulator are segmented through using a genetic algorithm-based optimal entropy double-threshold method; as for gaps in the water drops, a closed region filling function is adopted; in order to improve the accuracy of identification, five parameters with the largest correlation with the hydrophobic grade are extracted, and a large number of water drop/water film images are acquired so as to train a BP neural network model, so that a 5-11-1 type BP neural network model for judging the hydrophobic grade of the contaminated composite insulator is obtained; and the model is adopted to process the water drop/water film images which are obtained on the basis of a water spray grading method, so that the hydrophobic grade of the composite insulator can be successfully determined. With the method adopted, the influence of the subjective factors of the spray water grading method can be eliminated, the accuracy of the identification of the hydrophobic grade of the contaminated composite insulator can be greatly improved; and an existing manual contamination test method for insulators can be further improved.