<p>Suction-induced shrinkage in drying fine sediments is a critical concern for geotechnical infrastructure with high clay content under climate stressors. However, characterizing internal deformation within fine sediments remains challenging. Particle tracking methods (PTMs) are effective for monitoring internal deformation of fine sediments by tracking markers embedded within the sediment; however, existing approaches struggle with a large number of markers and complex displacement fields. This study introduces a hybrid PTM that integrates the Local Relative Positioning-based (LRP) and the Artificial Neural Network-based (ANN) particle tracking method to monitor complex deformation within fine sediments. Micron-sized glass bubbles were embedded in saturated bentonite as tracking markers, and multiple CT scans were performed during the drying process to capture internal deformation before and after desiccation crack formation. Image processing techniques were applied to extract the location and morphology information of the glass bubbles. The proposed method first employs LRP to track markers based on the similarity of particle relative positions quantified at a local coordinate system. Then, LRP results are used to train a displacement trend model using the Random Forest algorithm, enabling the prediction of each glass bubble’s ‘potential’ location after sample deformation. Finally, the ANN further refines tracking by integrating the predicted displacement trend and links glass bubbles by minimizing discrepancies in particle location and morphology. The hybrid method obtains more accurate displacement fields with reduced noise and a much higher tracking ratio. The proposed method provides a more robust tool for analyzing the complex internal non-uniform deformation of fine sediments.</p>