Solar spectrum observation is one of the effective methods to study solar atmospheric phenomena. In this paper, a method of extracting and reconstructing solar spectral information based on principal component analysis (PCA) was proposed. Besides, the relation between the noise suppression degree of reconstructed data and the order of principal components was analyzed. In addition, the signal-to-noise ratio of the spectral line and the accuracy of the Doppler velocity measurement were calculated under different principal component orders. The results showed that after the feature information extraction, the reconstructed data greatly preserved the original spectral data, and their signal-to-noise was markedly improved, thus the Doppler velocity measurement accuracy of spectral line was significantly improved, and also the amount of data storage and transmission of the 3D spectral data were greatly reduced. This method can satisfy the releasing requirements of current data standard and scientific goals of the 1-meter New Vacuum Solar Telescope. This method also provide a reference for the spectral data processing of the under construction Fiber Arrayed Solar Optical Telescope and future Chinese Giant Solar Telescope.