在INS/CNS组合导航系统中，为了有效实时地用CNS的输出姿态信息校正INS的陀螺漂移，必须提高CNS中星图识别算法的运行速度。提出了一种改进的三角形星图识别方法，基三角形识别成功后，增加一颗附加检测星，计算出附加检测星和基三角形的三个星点的角距值。分析结果表明传统的星图识别算法在主频为600 MHz的DSP6414上运行需要300 ms左右的时间，算法改进后，只需40~80 ms，数据输出率提高到10 Hz。改变的算法在不存在伪星的情况下成功识别率近100%，存在伪星时，可以有效去除伪星。同时，解决了传统星图识别算法在嵌入式系统中实现难、耗时长的问题。该算法可以实时地校准惯导系统的误差。

In order to use effectively the output attitude information of CNS to correct gyro shift of INS in real time，the operating speed of star image identification algorithm in CNS must be improved in the INS/CNS combination celestial system. An improved triangle star image identification method is presented. After the base triangle is sucessfully indentified, an additional star is added and the angular distances of the star pair are calculated.The running time is about 300 ms on DSP6414 with the main frequency of 600 MHz by using the traditional star image identification algorithm. The time required is only 40～80 ms by the improved algorithm and the output rate of the data is improved to 10Hz. If there is no false star，the successful identification rate of the algorithm will reach 100% approximately. If the false stars exist，it can effectively remove them. At the same time，the problems，which it is difficult to realize in the embedded system for the traditional star image identification algorithm and the operating time is very long are solved. The improved algorithm can greatly improve the device performance and effectively calibrate the error in the inertial navigation system in real time.