In order to solve the ill-posed problem in 3D acoustic temperature field reconstruction, the paper proposed a new modified singular value decomposition (SVD) method.According to their reliability ,singular values were divided into three parts and got various degree of modification respectively. To verify the performance of the new algorithm based on the modified SVD method,two model temperature fields were reconstructed when the signal-to-noise ratio (SNR) of sound flight-time data was 50dB , 40dB and 30dB respectively.And the results were compared with those based on routine Truncated singular value decomposition (TSVD) and Tikhonov methods. Simulation results show that the new algorithm has higher precision, better anti-noise ability than the routine methods and it is more suitable for the complex temperature fields reconstruction, thus it is expected to be used for temperature field reconstruction on-line.