In the process of fluorescence molecular tomography reconstruction, there are different optical parameters in the reconstructed image for the fluorescent targets with the same optical parameters, the same embedded depth, but different volumes in the tissues. To solve this problem, we propose a novel fluorescence molecular tomography reconstruction algorithm based on volume compensation. We use an improved iterative self-organized data analysis techniques algorithm (ISODATA) to decide the center and the expectation number of the initial clusters and to make a cluster analysis of the pre-iterative reconstruction image. According to the sizes of different fluorophores clustered by the improved ISODATA, we design an operational method with volume weight coefficient based on logarithm to compensate nonlinearly the reconstructed optical parameters of different fluorophores. The simulation results show that our compensation method can effectively amend the reconstruction error resulted from the volume difference of the fluorophore.