Method for recognizing fog concentration of hazy image

A method for recognizing a fog concentration of a hazy image, the method comprising:

• inputting a target hazy image into a pre-trained directed acyclic graph (DAG) support vector machine to acquire a fog concentration of the target hazy image;
• wherein:
• the fog concentration of the target hazy image is represented based on a prebuilt multi-feature model, and a feature vector in the prebuilt multi-feature model comprises at least one of a color feature, a dark channel feature, an information quantity feature, and a contrast feature;
• wherein training the DAG support vector machine comprises:
• acquiring hazy images with different fog concentrations to build a training set;
• and
• inputting the training set into the DAG support vector machine, and supervising and learning the feature vector in a multi-feature model for representing the fog concentration of the hazy image by an S-directed acyclic graph support vector machine (DAGSVM) multi-classification algorithm based on a separability optimized DAG graph;
• wherein supervising and learning a feature vector in a multi-feature model for representing the fog concentration of the hazy image by an S-DAGSVM multi-classification algorithm based on a separability optimized DAG graph comprises:
• defining a separability between any two classes of samples in the training set, assuming any two classes of sample sets in the training set as {x1,x2, . . . xn} and {y1,y2, . . . , ym}, defining an inter-class distance dQ(exQ,eyQ) and an intra-class divergence σxQ of the two classes of sample sets in a feature space Q after mapping, and combining the inter-class distance and the intra-class divergence to define a separability SijQ=dQ(eiQ,ejQ)−β(σiQ+σjQ) between a class i and a class j, where β is a weight coefficient;
• inputting the training set into the DAG support vector machine, and combining various classes of sample sets in the training set in pairs to train two-class SVMs, k(k−1)/2 two-class SVMs for a k-class problem, and the classes correspond to different fog concentrations;
• calculating a separability between every two classes according to the separability between the class i and the class j to acquire a separability set S={sij},i≠j,i,j∈D, where sij=sji, and D is a class set in the training set and corresponds to a set of different fog concentrations;
• extracting a maximum separability max{sij} from the separability set, and using a two-class SVM corresponding to the maximum separability as a root node of the DAG graph;
• and
• assuming the two-class SVM of a current node as cij, and deciding which two-class SVM to be used in a next layer according to a result of classification of the two-class SVM of a current node.