Image Fog Density Recognition Method Based on Multi-Feature Model and S-DAGSVM

Aiming at the demand of intelligent image defogging systems for automatic recognition of fog density, this paper proposes an image fog density recognition method based on multi-feature model and S-DAGSVM. By analyzing the image characteristics of different fog densities, a multi-feature model based on the combination of four features of color, dark channel, information and contrast is constructed to characterize the image fog density, and the features are represented in the form of a histogram. Then, the S-DAGSVM algorithm is proposed to performs supervised learning on the combined feature vectors to realize automatic classification of the image fog density. The experimental results show that the proposed multi-feature model can characterize the fog density more efficiently. Compared with existing multi-class SVM algorithms, the S-DAGSVM algorithm has a higher classification accuracy of up to 96.19%, which has a good reference value for the realization of intelligent image defogging systems.