Algorithm 2: Bagging algorithm in the proposed model

Input: Datasets $(X,Y)=\{(x_1,y_1),(x_2,y_2),(x_3,y_3),...,(x_m,y_m)\}$, $x_j\in X,x_j$ is an image, $y_j\in Y,y_j$ is the label of $x_j$: Base learner algorithm: Iterations n. Process: for $i=1,2,3,\cdots ,n$ Sampling randomly from training set X using bootstrapping and $X_i$ can be obtained, $X_i$ is a subset of the training set X: Training base learner $B_i$ with RL ondataset $X_i$; end for Output: The results of each base learner $B_i$ are combined into a final result by a plurality voting strategy, $F\left(x\right)=argma{x}_{y\in Y}{\sum }_{i=1}^n\prod (f_i\left(x\right)\neg y)$