colmap / colmap

COLMAP - Structure-from-Motion and Multi-View Stereo
https://colmap.github.io/
Other
7.73k stars 1.53k forks source link

The problem of sparse reconstruction #2902

Open MikeAiJF opened 1 day ago

MikeAiJF commented 1 day ago

Why do I use clomap for sparse reconstruction of fixed shot data, but the results are very poor, outputting two folders, 0 and 1, in sparse? Why is this? Is Colmap not friendly to fixed lens shooting data?

tsattler commented 1 day ago

If Colmap reconstructs multiple models instead of a single one, then it is unable to connect some images (typically due to a lack of feature matches). You can inspect the feature matches found between images in the gui.

Capturing more images in these regions can help prevent the split. Also, specifying the intrinsic calibration of the images and enforcing that a single camera model is used for all images (if they were taken by the same camera without changing the zoom) can help make the reconstruction process more robust.

MikeAiJF commented 22 hours ago

如果 Colmap 重建多个模型而不是单个模型,则无法连接某些图像(通常是由于缺少特征匹配)。您可以在 GUI 中检查在图像之间找到的特征匹配。

在这些区域捕捉更多图像有助于防止分裂。 此外,指定图像的固有校准并强制对所有图像使用单个相机模型(如果它们是由同一台相机拍摄且未改变变焦)有助于使重建过程更加稳健。

Thank you for your answer. My data is for a single purpose, but I found that the latest Colmap version on the virtual machine doesn't seem to work as well as Windows 3.6 or 3.7. Why is this?, What do single model and multiple models refer to?