Summary: In this article, we consider the estimation of possibly multiple change points in multivariate regression models with structural changes. A salient feature of the methods is that the dependence structure of the error terms and the regressors can be as weak as that of \(\mathcal{L}^2\)-mixingale arrays of size \(-1/2\). Further, we also provide some numerical simulations and a real data application to illustrate the efficiency of the proposed methods.