We’re working on generating a high-resolution 3D model of a tall, white high-rise building with a lot of glass. The uniform color of the facade and reflective windows are creating major challenges for photogrammetry, since Pix4D struggles to extract and match tie points on featureless or reflective surfaces. Our initial capture included both nadir and oblique flights around the structure, producing 684 images, but only about 76% calibrated successfully, leaving many facade shots uncalibrated (red). We’ve tested different mission types in DJI Pilot 2 (grid, spiral) and adjusted overlaps, but the model still separates into multiple blocks instead of a single, consistent reconstruction. We also tried adding manual tie points to strengthen weak links between image sets.
To address this, we flew a dedicated close-range facade mission to boost image detail and tie point density in the problem areas. The plan now is to merge this new dataset with the original capture in Pix4D by adding identical Manual Tie Points (MTPs) across both sets. Our hope is this will anchor the close-up shots to the wider model, improve calibration in the white, glass-heavy sections, and resolve the multiple block issue we’ve been seeing. I am just hoping for any tips or advice on accomplishing this challenging task.