The coherent structure characteristics of atmospheric turbulence are investigated based on frequency domain wavelet analysis method. The data of wind velocity fluctuation over the ground are analyzed with Mexican Hat wavelet function. According to the maximum energy principle, the principal scale of coherent structure is identified and the coherent structure waveform is extracted by reconstruction formula. Results show that the reconstructed coherent waveform is a signal with quasi-periodic motion, which reflects a middle scale movement of the existing original pulse signal, and their waveforms coincide well with each other in the changing trend. Time and scale maps of wavelet coefficients can well explain the multi-scale and intermittence characteristics of turbulence, and reflect the evolution of turbulence coherent structure. During the analysis period, statistical analysis on the main scale of wind speed coherent structure is carried out, and it’s found that the main scale of coherent structure satisfies the normal distribution approximately.