Integral time is the key factor to determine infrared focal plane arrays (IRFPA) imaging quality. In order to solve the problem that a infrared measurement system can’t real-time determine the optimal integration time, an IRFPA integral time adaptive predictive method based on radiation calibration and human visual characteristics was brought forward. Supposing the target and background in two consecutive frames don’t change, the information of target and background during the other integral time can be computed by combining with the results of prior radiation calibration and the information of target and background. Then, the optimal integral time can be determined automatically by the judgment based on human visual characteristics and tracking criterion. Experimental results show that the method can rapidly predict the optimal integral time, and make the image in the middle of the IRFPA response linear region. The method can not only ensure the image quality but also be conducive to the target initial capture and stable tracking.