Research On Image Classification Method Based On Broad Learning System For Limited Number Of Labeled Sample Conditions

Image classification is one of the research hotspots in the field of computer vision and pattern recognition,which has important theoretical significance and research value.Traditional image classification methods often rely on a great quantity of labeled samples.However,in practical application,it is time-consuming and expensive to obtain a great quantity of labeled samples.When the labeled samples are limited,it is difficult to obtain ideal classification results by using the traditional model.In order to reach the requirements for classification accuracy,efficiency and generalization ability,this article focuses on the characteristics of limited label samples and studies image classification methods based on the broad learning model.The main work includes:1.Aiming at the problems of low classification accuracy,poor stability and weak generalization ability caused by a small number of labeled samples in image data,a semi supervised(FLAP-BLS)image classification method based on Fick’s rule aided propagation(FLAP)and broad learning system(BLS)is proposed.Firstly,FLAP is used to mark plenty of unlabeled samples by few labeled samples in order to obtain a large number of labeled samples and build the sample data matrix.Then an efficient incremental BLS without deep structure can effectively extract features from large-scale data,it is used to effectively classify the sample matrix.Finally,USPS,MNIST and NORB datasets are selected to validate the effectiveness of the FLAP-BLS.2.In order to solve the problem that traditional broad learning system cannot use a small number of labeled samples and a great quantity of unlabeled samples simultaneously,an image classification method based on manifold regularization framework and BLS(SS-BLS)is proposed.Firstly,the features are extracted from labeled and unlabeled data by building feature nodes and enhancement nodes.Then the manifold regularization framework is used to construct Laplacian matrix.Next,the feature nodes,enhancement nodes and Laplacian matrix are combined to construct the objective function,which is effectively solved by ridge regression in order to obtain the output coefficients.Finally,the validity of the SS-BLS is verified by three different complex data of MNIST,and NORB,respectively.3.Aiming at the problem of traditional broad learning system that it is difficult to use the source domain label information to assist in enhancing the classification effect of the target domain,a domain adaptive BLS model(DABLS)based on the manifold regularization framework and MMD is proposed,and DABLS is applied to image classification.Firstly,feature nodes and enhancement nodes are constructed to extract features from source domain and target domain data.Secondly,the Laplacian matrix is constructed by using manifold regularization framework to explore the manifold characteristics in the target domain data and reveal the potential information of the target domain data.Then,the source domain data and the target domain data are used to construct the MMD penalty term to match the projection mean between the two domains.Thirdly,the feature node,enhancement node,MMD penalty term and Laplacian matrix of target domain are combined to construct the objective function.Then,the output coefficient is obtained by solving the objective function by ridge regression.Finally,the validity of DABLS is verified by public image data set.The research results can provide new ideas and technical reserves for image classification under the condition of less labeling,and have certain theoretical significance and practical value.