Abstract: Objectives: Accurate localization of underground pipelines is essential for urban infrastructure management. MEMS-based inertial navigation systems (INS) in urban underground pipeline surveying instruments utilize the non-holonomic constraint (NHC) to enforce zero lateral and vertical velocity, effectively mitigating position drift and enhancing measurement accuracy. However, geometric discontinuities at pipe joints or weld seams may introduce transient mechanical shocks when the pipeline inspection gauge (PIG) passes through these locations. Such short-term lateral and vertical disturbances occurring at pipe joints violate the NHC assumption, and then degrade positioning accuracy. Existing approaches typically address this issue by detecting pipe joints from inertial measurement unit (IMU) data and subsequently disabling or down-weighting the NHC observations. The effectiveness of these strategies, however, depends strongly on reliable joint detection, which may be affected by noise, corrosion, or deposits inside pipelines. Enhancing positioning robustness without explicit pipe-joint detection therefore remains an important challenge in underground pipeline localization. Methods: To mitigate the influence of transient disturbances, an INS/odometer/NHC integrated incremental observation correction algorithm is designed. Instead of directly employing instantaneous velocity observations, odometer measurements and NHC constraints are integrated over time to construct a displacement increment observation vector within a predefined update interval. These observations are then incorporated into a displacement-based observation model within an error-state Kalman filtering framework. Through time integration, large instantaneous velocity disturbances are transformed into small displacement increments with cumulative effects, thereby attenuating the influence of short-duration constraint violations on the overall state estimation. Results: Performance evaluation was conducted using 40 sets of field data collected from three pipelines with different lengths and environments. Two datasets correspond to above-ground pipelines with centimeter-level reference trajectories obtained from RTK measurements, while the remaining dataset corresponds to an underground pipeline with reference measurements provided by a high-accuracy PIG. Compared to the classical velocity observation model, the incremental model produces significantly smoother and more accurate measurement trajectories. Statistical analysis of the root mean square (RMS) values from 40 datasets shows that, for the planar trajectory error, the dispersion decreases from 5.88 m to 3.71 m, representing a reduction of 36.9%. The average dispersion at pipe joints decreases from 1.92 m to 0.66 m, corresponding to a reduction of 65.6%. In addition, the root mean square error (RMSE) decreases from 1.87 m to 1.43 m, indicating an improvement of 23.5% in planar positioning accuracy. These results highlight the superior robustness of the proposed model against short-term impact disturbances encountered during pipeline surveying. Conclusions: The displacement increment-based correction algorithm effectively mitigates the pipe-joint disturbance effects on positioning accuracy. The proposed method requires no additional hardware or tricky pipe joints detection, which ensure strong practicality for engineering applications in inertial pipeline surveying. The incremental observation concept may also may provide a useful reference for enhancing navigation robustness in environments subject to short-duration impulsive disturbances.