In order to solve the problem of intra pulse modulation type recognition in radar reconnaissance, the efficient recognition of NS, LFM, PSK and FSK modulation signals is studied. By improving the instantaneous autocorrelation algorithm, the autocorrelation function of radar radiated signal is obtained on the premise of suppressing phase ambiguity and noise interference, and the instantaneous frequency of signal is obtained by moving average sampling. Four different modulation types of signal frequency graph characteristics are analyzed. The self-correlation coefficient, the number of peak jumps and the second-order mean difference characteristics of frequency graph are extracted as recognition feature vectors, and an efficient intra-pulse modulation recognition algorithm is constructed. The simulation results show that the algorithm can distinguish 4 modulation types effectively under the condition of low signal-to-noise ratio. When the signal-to-noise ratio is higher than 9 dB, the recognition probability can reach more than 97%.