For the finite-time tracking problem for heterogeneous second-order nonlinear multi-agent systems, a fully distributed consensus control method is proposed to eliminate the dependence of the global information, including the upper bound of errors and the eigenvalues of the Laplacian matrix. Based on local information only, a finite-time consensus protocol is designed. To simplify the stability analysis of the time interval between two adjacent switching times, a distributed switching mechanism of the adaptive gains is proposed with the gains regulated as piecewise constants. By seeking a contradiction, it is proved that the multi-agent systems must reach finite-time consensus and the adaptive gains are bounded. A simulation example is provided to illustrate the effectiveness of the proposed fully distributed consensus protocol.