In this paper, an efficient algorithm is proposed to the angular-diversity radar recognition of ships with noise effects taken into consideration. The goal is to identify the similarity between the unknown target ship and known ships. The content of this paper is divided into two parts. In the first part, the angular-diversity radar recognition of ships is given by transformation based approaches, i.e., the principal components analysis (PCA), with noise effects taken into consideration. The goal is to identify the similarity between the unknown target ship and known ships. In the second part, the linear discriminant algorithm (LDA) is utilized to increase the recognition rate. Initially, the angular-diversity radar cross sections (RCS) from a ship are collected to constitute RCS vectors (usually large-dimensional). By changing the elevation angle or the ship type, different RCS vectors are obtained to produce a high-rank covariance matrix. By choosing some of the largest eigenvalues and their corresponding eigenvectors, all the RCS vectors are projected onto the eigenspace (usually small-dimensional). Similarity between the unknown target ship and known ships can be identified in the eigenspace with high recognition rate. This will reduce the complexity for radar recognition of RCS characteristics from ships. However, the separating ability for such an elementary recognition is usually poor. This poor separation of radar target recognition will make the prediction results unreliable. The PCA gives the major features for the projected data of the two classes. While the LDA gives the best separation for the projected data of the two classes. To enhance the separating ability of radar target recognition, the projection features on the PCA space are further projected onto the LDA space and the recognition is performed on the LDA space. Our simulation shows that the separating ability for RCS based recognition of targets is greatly increased by using the LDA in the radar recognition process. In addition, the use of LDA in the recognition process increases the ability to tolerate noise effects. This study will be helpful in many applications of radar target recognition.