In infrared detection, the quantitative analysis is usually based on the relationship between the specified characteristic time and the defect depth. Using a stainless steel plate with six flat-bottom holes as an experimental sample, the defect depth is calculated by using a PSDT method which need not to choose any reference. When the time of the second-order derivative of logarithmic temperature peak is taken as the characteristic time, the defect depth is measured according to the proportion of the characteristic time to the square of defect depth. The data calculation for any point can be implemented with a VC program. Firstly, the data is fitted by using the least square polynomial. Then, the second-order derivative peak time of the logarithmic temperature-logarithmic time is calculated. The result is displayed precisely in figures and the defect depth is automatically measured finally.