to improve identification speed of star images which are taken by drif t-scan CCD and realize real-time orbit determination of synchronous satellite,a fast star ima ge identification algorithm based on DSP is proposed.The big data is calculated on the platform o f embedded DSP, on this basis,the twice match method which is an optimized triangle identificat ion method is proposed to accelerate the calculation speed.The main steps of the new method a re that selecting reference stars from the star catalogue as the form of interval to as the first matching reference stars,then calculating the CCD negative model based on first matching outcomes, the CCD negative model represents the relationship between background stars and referenc e stars in catalog, in the end,matching all stars based on the CCD negative model.The experiment r esults of star images whose resolution is 1528×1528show that because of introducing DSP and improving identification method,the identification speed is become faster in a great part .The processing time taken by new method is only 50% than that of traditional triangle identifica tion algorithm. Meanwhile the identification rate is up to 98.4%.The outcomes of the experience s validate the advantages of the new method on aspect of real time capability an d identification rate.