Temperature plays a crucial role in influencing the mechanical properties of materials. Accurately measuring the temperature of devices is essential for understanding the evolution of their mechanical properties under stress and evaluating their health and lifespan. However,traditional methods encounter challenges in measuring transient temperatures and lack sufficient time-resolution capability,particularly when it comes to the rapid temperature changes at the solder interface during the switching process of power devices. In this paper,based on the close correlation between the intensities of the characteristic spectral lines of the laser-induced elements and the temperatures,a method of measuring the surface temperatures with the time-resolved capability of the order of microsecond is proposed,and a quantitative relationship between the surface temperatures of the sample and the spectral characteristics is established. The findings demonstrate that an increase in the surface temperature of the material results in enhanced intensity and signal-to-noise ratio of laser-induced plasma spectra. This enhancement is influenced by the spectral acquisition delay and gate width. To establish a quantitative relationship between surface temperature and spectral properties,back propagation-artificial neural network (BP-ANN) and partial least squares (PLS) are employed for fitting and calibration. The fitted models can achieve linear correlation coefficient indexes exceeding 0.99. Notably,the BP-ANN fitted model exhibites a small fitting bias,with a root mean squared error (RMSE) of 2.582 and a correctness rate of 98.3%. The method provides an effective means for transient temperature measurement of objects and gives a strong support for the assessment of the health status of the soldering interface of power devices.