The autocorrelation characteristics of the super-Gaussian optical pulse and their variations with the edge sharpness parameter, linear chirp parameter, noise-pulse and random noise are investigated by use of numerical analysis method, and compared with the relevant parameters of the original pulse. A useful method of filtering random noise is given and validated by experiments. When the edge sharpness parameter increases, the curve of intensity autocorrelation becomes narrow, and the autocorrelation spectral curve broadens without chirp, noise-pulse and random noise. When the linear chirp parameter increases, the temporal waveform of the pulse and the curve of intensity autocorrelation keep unchanged; the curve of pulse spectra broadens and the oscillatory structures appear at its edges and increase; the autocorrelation spectral curve broadens and its edges are still smooth. The variation of pulse autocorrelation characteristics is complex when noise-pulse or random noise exists.