Abstract: Bridge foundation scour is a major cause of bridge collapse.Existing indirect scour detection methods using moving vehicles are limited by interference from road surface roughness.To address this issue,this paper proposes a scour detection method based on the residual contact-point response of a two-axle test vehicle.The method eliminates roughness effects and reconstructs high-fidelity bridge mode shapes,thereby improving the robustness of scour identification.First,a coupled vehicle-bridge-pier model considering scour effects is established,where scour is represented as a local reduction in foundation stiffness.Second,a back-calculation procedure extracts the residual contact-point response from vehicle accelerations.A spatial differencing mechanism between the front and rear axles effectively removes road roughness interference.Then,zero-phase filtering and the Hilbert transform are integrated to reconstruct high-fidelity,high-resolution mode shapes.On this basis,two damage indicators are constructed:the difference in mode shape curvature (DMSC) for scour localization and the root-mean-square (RMS) deviation for scour severity quantification.Numerical results show that DMSC exhibits a distinct peak at the scoured pier,enabling accurate scour localization.RMS increases monotonically with scour severity,allowing effective quantification.The method maintains good robustness under rough pavement conditions and can be extended to multi-span continuous beam bridges with unequal spans.This study provides a new technical approach for rapid identification of bridge foundation damage using moving vehicle responses.