Corrosion is a leading cause of pipeline failures, responsible for up to a third of reported incidents. Given the extensive length and limited accessibility of pipelines, operators rely on frequent in-line inspections (ILIs) to detect and quantify corrosion defects. Feature matching between successive ILIs is therefore utilized to align and compare this data, enabling the identification of corrosion evolution and behavior. However, measurement uncertainties, nonuniform corrosion growth, and spatial interactions between adjacent defects pose significant challenges in achieving accurate matching. This study proposes a two-phase feature matching model designed to address these limitations. The first phase performs Iterative Closest Point (ICP) alignment with a context-aware nearest neighbor selection strategy based on Directional Epsilon Neighborhood Clustering (DENC) to isolate stable feature pairs and minimize ambiguous associations in densely clustered defect regions. The second phase applies a novel proximity–overlap-informed correspondence optimization using linear programming to identify matches and outliers by jointly considering feature positioning and geometric attributes. The model’s effectiveness is evaluated on a 1116 m subsea pipeline segment involving two consecutive inspections reporting 1305 and 1491 features, respectively. Compared to three state-of-the-art models, the proposed approach achieves a recall, precision, and F1 score of 99.2%, demonstrating substantial improvements in accuracy, stability, and robustness to inspection and corrosion-related uncertainties. These results confirm the model’s ability to address critical limitations in existing approaches and underscore its potential to enhance pipeline integrity assessments.