Abstract: In order to reconstruct complicated temperature fields more accurately by acoustic CT,a new reconstruction algorithm based on Markov radial basis function and Tikhonov regularization is proposed and named as MTR algorithm. With the algorithm,the acoustic velocity field in a medium is approximated by a liner combination of Markov radial basic functions,the acoustic travel-times over multi-paths and the Tikhonov regulation are used to reconstruct the acoustic velocity distribution,and then the temperature distribution is calculated by using the relationship between acoustic velocity and temperature.The temperature field models with one hot spot,three hot spots and five hot spots are reconstructed by using simulation data.Reconstruction results show that MTR algorithm can reconstruct the hot spot temperature,especially the hot spot position accurately.An experiment system for temperature distribution measurement by acoustic tomography is developed.The capability of acoustic tomography to detect a hot spot created by electric heaters in an experimental silo filled with 1200 kg soybeans is tested by using MTR algorithm.In the reconstruction temperature field,the hot spot position can be determined accurately and the temperature error of the hot spot is 1.3%.It is thus clear that MTR algorithm has a good capacity for reconstructing complex temperature fields, and can be expected to be used in temperature monitoring for actual stored gains.