The existence of porosity defects on the weld surface reduces the effective cross-sectional area of the workpiece and the ability of the workpiece to resist external loads, which leads to workpiece fracture in serious cases. Therefore, a multi-defect detection method for weld surfaces based on eddy current pulse thermography technology is proposed. First, a new electromagnetic sensor structure is adopted to detect carbon steel defects with different diameters and depths through the principle of eddy current pulse thermography, and the temperature signals of the defect and non-defect areas in the image sequence are analyzed. To improve the sensitivity of the detection system, principal component analysis (PCA) is used to reconstruct the image sequence and enhance the defect features in the original image. Finally, the proposed method was verified experimentally. The experimental results show that this method can reduce the influence of the weld edge effect, realize large-area detection of weld surface defects, and provide an open field of vision for infrared thermal imaging.