In the complex electromagnetic environment, the feature parameters could be classified with favorable separating degree in high Signal-to-noise ratio (SNR) by conventional gray relation algorithm. In low SNR, however, it will be existed overlapping phenomenon and even hard to select standard sample values of unstable parameter. Aiming to these issues, this study does further research on the classifier design method based on gray relation algorithm. By introducing the mean samples gray relation, the selection of standard sample value is improved. Moreover, the features are weighed according to the surplus degree which increases the adaptive ability of the gray relation algorithm. Meanwhile, the paper used the interval relation classifier based on adaptive entropy weight to classify the features with overlapping properties. It overcome the problem of selecting standard samples and had some adaptive ability to some extent. Comparing with other algorithms, it has the best classification effect.