Abstract: Objective Buried long-distance pipelines operating in complex geological environments are often subjected to geological movements, landslides, soil corrosion, and other external disturbances, which may induce deformation and lead to structural failure. Among various pipeline components, bends represent geometric discontinuities that are prone to stress concentration and thus constitute high-risk locations for fatigue failure, local buckling, and stress corrosion cracking. Accurate identification of bend locations and spatial configurations is therefore essential for pipeline integrity assessment and safe operation. However, existing in-line inspection technologies still show limitations in comprehensive bend detection, type discrimination, and spatial geometric characterization, particularly for pipelines in geologically complex regions. Methods To address these challenges, this study proposes a bend identification and spatial feature analysis method based on inertial measurement unit (IMU) data and odometer readings acquired by in-line inspection tools. Pipeline attitude is obtained through inertial navigation computation, and, combined with odometer information, enables precise bend localization, type classification, and reconstruction of the local three-dimensional trajectory. A small-diameter gathering pipeline of approximately 4.4 km in length is selected as a case study, and the proposed method is validated against magnetic flux leakage (MFL) inspection results. Results The proposed method identified a total of 151 bends, whereas MFL inspection detected 137 bends, with 128 matched locations, yielding a concordance rate of 93.43% when using MFL as the reference. In addition, the algorithm successfully identified 22 bends that were missed by MFL inspection, all of which were confirmed to exhibit typical bend geometries through 3D trajectory reconstruction. Conversely, bends identified solely by MFL were mostly located near girth welds and did not exhibit significant continuous curvature. Conclusion Overall, the IMU-based bend identification method demonstrates clear advantages in detection completeness and spatial geometric representation. The proposed approach provides reliable data support for pipeline integrity evaluation and safety management.