In this paper, we mainly deal with the image-based visual servoing of nonholonomic mobile robots with onboard camera. The depth-independent interaction matrix is firstly applied and extended in our work, where normalized image coordinates are utilized to estimate unknown information coupling the orientation of mobile robot and the translational component between the feature points and the mobile robot. Then, a image-based adaptive controller is proposed which could handle with time-varying feature depth. Lyapunov theory is carried out for the stability analysis. Both simulation and experiment results are collected to investigate the feasibility of the proposed approach.