To address the obstacle-avoidance path-following control problem of unmanned surface vehicles subject to time-varying disturbances and dynamic uncertainties, a modular integrated planning-guidance-control framework is proposed. In the planning layer, a visual projection technique combined with an obstacle-avoidance path optimization mechanism (OAPOM) is employed to generate safe paths that satisfy maneuverability requirements. In the guidance layer, to handle the underactuated nature of the USV and compensate for sideslip, a novel fix-time predictor line-of-sight (FxTPLOS) guidance scheme is developed together with a time-varying sideslip predictor. Compared with existing studies, the proposed method establishes a three-module design framework, in which the OAPOM-based obstacle-avoidance path is seamlessly integrated with the FxTPLOS guidance law, while the guidance layer provides desired heading and speed signals for the control layer, thereby forming an integrated path-planning-guidance-control architecture. Based on Lyapunov theory, it is shown that all signals in the closed-loop obstacle-avoidance path-following control system are uniformly ultimately bounded. High-fidelity simulations verify the robustness, path feasibility, and practical applicability of the proposed scheme.