The LTE-based V2V communications is considered to be the key technology to improve the performance of the intelligent transportation system. In order to allocate LTE resources efficiently on the scenarios of large-scale LTE-V2V networks, a novel clustering algorithm based on the similarity of mobility is proposed in this paper. We present a new definition — degree of link dependence to measure the similarity of mobility and a novel maintenance scheme which can reduce the rebuilt cost. In the proposed algorithm, vehicles are grouped in one cluster only if they are with similarity mobility. The nodes selected as heads will not be reselected every cycle but keep their roles on the straight road, and only under some certain circumstance the heads list will be updated; but the ordinary nodes always continuously execute attaching-head process by themselves. In addition, under the topological structure of cluster, we devise an alarm information forwarding scheme which can distribute the alarm packets efficiently and reliably under half-duplex communication mode. The simulation results represent the performance of the algorithm in terms of the cluster head lifetime and the average reattaching-head times of ordinary nodes.