To avoid the ill-posedness in the inverse problem of bioluminescence tomography, a moment searching algorithm fusing the finite element method (FEM) with the moment concept in theoretical mechanics is developed. In the algorithm, the source's information is mapped to the surface photon flux density by FEM, and the source's position is modified with the feedback through the algorithm of barycenter searching, which makes full use of the position information of the photon flux density on surface. The position is modified in every iterative step and will finally converge to the real source's value theoretically.