[Purposes] This paper aims to explore the influence of noise generated by experimental environments and experimental equipment on the positioning accuracy of defect imaging by Lamb wave signal. [Methods] In order to optimize the sensitivity of the signal difference coefficient to the damage, the Lamb wave signal sampled in the plate was decomposed and reconstructed by a wavelet packet to get the reconstructed Lamb wave signal in the excitation frequency range by combining simulation and experiment. [Findings] After the Lamb wave signal is reconstructed, the probability value of the real damage location of the plate is higher, and the range with a high probability value is closer to the damage location. [Conclusions] After the Lamb wave signal is decomposed and reconstructed by a wavelet packet, compared with the damage probability imaging of the original signal, it is found that damage probability imaging considering reconstructed signals can effectively locate defects in the plate, with improved positioning accuracy and contrast ratio of imaging.