Olli Huttunen has released Point Cloud Merger for Blender, an add-on that merges multiple COLMAP reconstructions into a single training package. A site covered from the ground and again from a pole or a drone produces two independent reconstructions, each with its own scale, origin and orientation, and a gaussian splatting trainer accepts one. Point Cloud Merger performs that reconciliation inside the Blender viewport.
Huttunen's 360 Extractor plugin for Blender reached version 1.2.1 in February with support for full 360 image sequences, and Dioramix, his browser based presentation editor announced in April, loads PLY, SPZ and SOG captures for annotated walkthroughs.
Huttunen describes the add-on as a COLMAP point cloud and camera pose editor for Blender. Import brings in as many COLMAP reconstructions as needed, each arriving as a colored point cloud with its cameras correctly posed. Cloud orientation and point display can be adjusted after import.
Alignment is done by picking a few matching points across two overlapping scans and snapping one onto the other with a similarity fit, with cropping available for anything outside the area of interest. A similarity fit solves rotation, uniform scale and translation from the picked correspondences, which is what two reconstructions of the same site need when neither carries metric scale.
Export writes the combined result as one COLMAP package and applies the Blender to Gaussian Splatting axis conversion automatically. Huttunen names Postshot and LichtFeld Studio as targets, along with any other COLMAP compatible 3DGS trainer.
A single tutorial accompanies the release, a 17 minute walkthrough of import, orientation, point cloud appearance and alignment that Huttunen labels Part 1.
A surveyor who has been running a ground level LiDAR pass and a pole mounted 360 pass as separate trainings, then picking whichever came out better, can train both as one dataset. Point Cloud Merger is a 51 Kb ZIP, requires Blender 4.2 or higher, and is sold through Huttunen's ToppiNappi storefront under a royalty free license at a fixed $30.33.
