Abstract: In order to survey the radiation of nuclear power plants, an algorithm for reconstructing radiation map was built based on an intelligent inspection robot system. A set of discrete radiation data with location information was collected through simultaneous localization and mapping method. Based on the assumption of single point radioactive source, an improved particle filter algorithm was developed to estimate the parameters of the radioactive source and reconstruct the radiation distribution. Finally, the distribution was combined with the scene map for a visual presentation. The effect was verified through the comparison between measured and simulated values. The results show that in the scene with a single point radiative source the average relative error is about 16% and it works well even if the input radiation data is relatively sparse.