Performance comparison of target classification in SAR images based on PCA and 2D-PCA features

Feature extraction is an important step for target classification in SAR images. Principal component analysis (PCA) is common in pattern recognition, and has been used widely for target classification in SAR images. In order to utilize PCA, two-dimensional image has to be arranged to an observation vector. However, two-dimensional PCA (2D-PCA), which is developed from PCA, can extract features from two-dimensional SAR image directly. Although 2D-PCA is consistent with PCA in theory essentially, which represents original data by extracting principal components with high variance values by linear transformation, they perform distinctly due to the difference of data processing methods. Based on the theoretical analysis and classification experiment using MSTAR data, this paper compares PCA and 2D-PCA systematically and roundly.