To improve the remote sensing image classification accuracy by incorporating labeled and unlabeled samples

this paper proposed a new manifold learning method

called semi-supervised manifold discriminant embedding (SSMDE). This method constructs two relational graphs through data point labels: A within-class graph and a between-class graph are used to encode the class relation information indicated in the labeled data points

and two weighted matrices are constructed based on the two graphs. Then

we utilize the labeled and unlabeled data points to construct the total scatter matrix to describe the information of all data points. Finally

the projection matrix of SSMDE can be found by solving an optimization problem. SSMDE method can not only take into account the discriminant information of labeled data

but also preserve the global structure information of all data points. The experimental results on both synthetic and remote sensing images show that the proposed method can achieve the classification accuracy rate (92.32%) and the error between the classification results by SSMDE and the Government Statistics are less than 5%

which demonstrates the effectiveness of SSMDE.