Cracks are one of common defects in steel materials.In this paper,the quantitative detection of steel surface cracks is studied using pulsed eddy current thermography.The 45# steel with slot cracks is inductively heated,and the temperature distribution and temperature response around cracks are analyzed.The results show that,compared with temperature rise,the quantitative relationship between the shape of temperature response curves and crack depth is clearer,and the shape of curves can be characterized by the phase of Fourier integral of temperature.The crack geometric profiles are reconstructed by the least squares support vector machine(LS-SVM).Results are compared when the input data of the LS-SVM are phase profiles and temperature profiles respectively,and it reveals that the phase profiles are more suitable for the quantitative detection of surface cracks.