Multispectral radiation thermometry(MRT) is one of the methods for non-contact temperature measurement. The present MRT could invert only one approximate temperature value, and the actual temperature distribution could not be given. In the present article principal component analysis(PCA) was applied to solve the problem based on Planck’s laws. For example, in the laser damage target experiments, input laser pulse energy is 34.2 J, the inversion temperature value is 2 700 K from the thermal radiation spectrum. The fitting relative deviation is 0.5%, and when the temperature distribution is from 2 600 to 2 800 K, the fitting relative deviation is reduced to 0.13%. Temperature distribution inversion was shown based on PCA for the first time, and the measurement accuracy of non-contact temperature was improved.