The traditional reconstruction method of Cerenkov luminescence tomography (CLT) has a relatively low reconstruction accuracy. In this paper, we propose a finite-element-subdivision-based local connection network (FES-LCN). The network consists of a fully connected subnetwork and a node-connected subnetwork, in which the latter is based on finite element subdivision and used to predict the residual error between the initial reconstruction result and real light source. The physical simulation experiment and digital mouse experiment are designed to test the performance of the proposed network. The results show that the FES-LCN has good stability and accuracy.