As one of the important structures of composite materials, polymethacrylimide (PMI) foam sandwich structure is prone to debonding and delamination defects during manufacturing and service, which seriously affects the mechanical properties of materials. Therefore, it is necessary to detect the defects of foam sandwich structure. A linear frequency modulation infrared thermal wave nondestructive testing system was built to study the correlation between the geometric characteristics (diameter and depth) of the defect and the surface temperature signal and its influence on the detection effect. The image sequence is processed by principal component analysis, discrete Fourier transform, thermal wave signal reconstruction, and other algorithms. Studies have shown that the larger the defect diameter, the shallower the depth, easy to detect defects, detection effect is good. The detection system can effectively detect and identify debonding and de-lamination defects. The thermal wave signal reconstruction method is superior to the principal component analysis method and the discrete Fourier transform method in the signal-to-noise ratio index, which improves the quality of the image and is conducive to the effective identification of the image surface defect information.