Speckle elongation phenomenon can be observed when rough surface is illuminated by a polychromatic laser beam. Based on this dependence the surface roughness can be determined, in which the influence of transverse characteristic of rough surface can be neglected when certain conditions are satisfied. Through numerical simulation of polychromatic speckle pattern of random rough surface, by application of the local autocorrelation function of polychromatic speckle pattern using spatial average, the dependence of average speckle elongation ratio 〈χ〉 on root-mean-square σh of surface profile amplitude is studied, and the influences of experimental factors are investigated, such as the wavelength combination of laser beams, the cell size and the resolution depth of image acquisition device on measure result. The results show that describing the polychromatic speckle elongation effect by autocorrelation function using spatial average instead of ensemble average is effective, and to obtain enough precision of surface roughness measurement, appropriate wavelength combination of laser beams and appropriate cell size of image acquisition device should be chosen.