Abstract: Lamb wave has been extensively used in nondestructive testing of composites due to its wide detection range and high sensitivity to defects. However, strong noise environment makes it difficult to extract effective signals, which affects the accuracy of damage location determination. To solve this problem, a damage location imaging method based on box-counting dimension and Lamb wave tomography technology in strong noise background was proposed. Firstly, the propagation characteristics of Lamb wave before and after damage of composite plate were analyzed by simulation. Circular sensor arrays were uniformly arranged on the carbon fiber reinforced polymer(CFRP) composite plate to simulate the real damage by changing the strain field of the structure with bonded mass blocks. Secondly, each sensor in turn acted as an exciter to generate Lamb wave. Other sensors collected the response signal without damage, and used wavelet transform to denoise the signal. Different levels of white noise were added to the denoised signal to realize noise interference. Finally, the damage factor was determined by calculating the difference of non-destructive signal by box-counting dimension, and the damage location imaging was realized by probability imaging algorithm. The experimental results show that the maximum and average errors of single-damage and multi-damage imaging locations are 11.18 mm and 6.88 mm respectively in strong noise environment. This method does not need signal denoising technology, and avoids the extraction process of complex reflection signals when multi-damage occurs. It has great potential in damage location and identification of composites under strong noise.