Research on Error Point Deletion Technique in Three-Dimensional Reconstruction of ISAR Sequence Images

Three-dimensional reconstruction using a two-dimensional inverse synthetic aperture radar (ISAR) faces dual challenges: geometric distortion in initial point clouds caused by accumulated feature-matching errors and degraded reconstruction accuracy due to point cloud outlier interference. This paper proposes an optimized method to delete the error points based on motion vector features and local spatial point cloud density. Before reconstruction, feature point extraction and matching for ISAR sequence images are performed using Harris corner detection and the improved Kanade-Lucas-Tomasi (KLT) algorithm. To address the issue of mismatched points, a method based on motion vector features is proposed. This method applies the dual constraints of motion distance and direction thresholds and deletes mismatched points based on local motion consistency. After point cloud reconstruction, a clustering method based on local spatial point cloud density is employed to effectively remove outliers. To validate the effectiveness of the proposed method, simulation experiments comparing the performance of different approaches are conducted. The experimental results demonstrate the effectiveness and robustness of the proposed method in the 3D reconstruction of moving targets.