Aiming at the LAMOST project, the largest sky survey project implemented in China, a novel method for automatic recognition of emission-line stars (ELS) based on detection of spectral lines is proposed. This method consists of three main steps: 1) the integral information of spectral lines is calculated by obtaining the correlations of the features of spectral lines; 2) the stellar Balmer lines are detected based on the features of stellar spectral lines; 3) the ELSs are automatically recognized by using the threshold of the correlations and the results of the stellar Balmer lines. In the experiments with all observed spectra from the sloan digital sky survey (SDSS) DR4, 242 peculiar celestial bodies with the stellar emission lines are obtained. According to the Simbad reference database, they are the celestial bodies include ELSs, cataclysmic variable stars (CVS), and unknown peculiar celestial bodies, etc.. The extensive experiments with observed spectra show that the proposed method can automatically recognize ELS effectively.