3D UAV Navigation Map Reconstruction With Selective Image Matching

The patent divides the image set into multiple groups, where each group contains a subset of images. Feature point matching is performed within each group independently, and then results are integrated. This segmentation approach reduces the total number of pairwise comparisons from O(n²) to O(k·m²) where k is the number of groups and m is the average group size, significantly reducing time consumption while maintaining matching completeness through result integration.

The patent performs preliminary feature point extraction and image preprocessing before the matching step. By pre-processing images to enhance features and pre-extracting keypoints, the actual matching process becomes more efficient. This preliminary action ensures that when images are matched, the feature points are already optimized for comparison, reducing the time required for the matching operation itself while maintaining accuracy.

The patent uses two-dimensional images as copies or projections of the three-dimensional scene, and through computational algorithms (SFM and MVS), reconstructs the three-dimensional structure from these 2D copies. This approach replaces expensive laser scanning equipment with ordinary cameras, as the 2D image copies contain sufficient information for 3D reconstruction when processed through the proposed methodology.

The patent replaces mechanical laser scanning systems with a computational imaging approach. Instead of using physical laser scanners to directly measure 3D coordinates, the system uses 2D image capture followed by computational algorithms (feature matching, SFM, MVS) to reconstruct 3D models. This substitution reduces device complexity while achieving comparable or superior precision through software-based reconstruction.