The guided wave EMAT detection of large metal plate components has the problems of low resolution and poor SNR of the EMAT, which is challenging to be used for small crack detection. A multi-physical field finite element model of Lamb wave EMAT detection process based on the Barker code pulse compression technology is established to solve this issue. Taking a 5. 6 mm thick steel plate with prefabricated cracks as the detection object, the effects of permanent magnet width and height, meander coil turns, the bit length of Barker code sequence, and subpulse duration on EMAT detection echo are studied by simulations and experiments. The beneficial effect of pulse compression technology in Lamb wave EMAT detection is evaluated. After optimization, results show that the SNR of the EMAT based on pulse compression technology is 9. 69 dB higher than that of the traditional tone-burst excitation method. A small crack 10 mm long and 0. 5 mm deep can be detected, and the SNR of the defect wave can reach 23. 47 dB. When the center frequency of the Barker code subpulse is 1 MHz, the Lamb wave packet after pulse compression presents mode separation. However, the SNR of the A0 mode defect wave can still reach 35. 23 dB, and it has essential engineering application value for improving the detection ability of the Lamb wave EMAT.