The invention relates to the field of reconstruction algorithms, in particular to a temperature field reconstruction algorithm based on a radial basis function and regularization. The problems that inthe prior art, the edge information of a temperature field is lacked, information fitting is not smooth enough and the precision is not enough are solved. The temperature field reconstruction algorithm comprises the following steps of A, dividing a to-be-measured region; B, calculating the flight time; D, calculating the reciprocal distribution; and E, calculating the temperature field distribution. According to the algorithm, global reconstruction is well achieved; good overall performance and smoothness are achieved, the reciprocal distribution of the two-dimensional plane ultrasonic soundvelocity is represented by the linear combination of the algorithm, good continuity is achieved, and the problem that the edge information of the temperature field in previous methods is lacked is solved; the temperature distribution of the edge of the temperature field is well reconstructed, the reconstructed temperature field is more smooth in distribution, and the temperature field distributionis closer to the real temperature field distribution; the problem of ill-conditioned matrix solving is solved, and the reconstruction precision is improved.