3D reconstruction implementation with OpenCV, which reconstructs a 3D object from a set of images. Key points are matched between the images either using a Lucas-Kanade tracker or a SIFT-implementation. The matched keypoints are used in a RANSAC algorithm to find camera matrices, and 3D positions are reconstructed. Finally, Delaunay triangulation creates a 3D mesh from the point cloud.
Can this work if the camera positions are random and unknown? e.g. if I take 5 or 6 photos from various angles around an object, can it detect and match features?
batlin 2 months ago
There is a PhotoScan product agisoft.ru/products/photoscan/ It automatically reconstructs 3D model with texture, exports it in necessary formats. And it works on real data, not only in laboratory ). Enjoy!
rodikov 7 months ago
Not bad. :)
Blooper1980 7 months ago
ANY downloadable app to test this at home?
pablovidaure 1 year ago
Hi, I have many problems in 3d reconstruction with OpenCV. Do you can help me? What instruction to use for find camera matrices, and 3D positions. I use cvTriangulatePoints and don't work very well. Or you know the camera parameters for this. Thanks.
tutv1980 1 year ago
It a great idea but really low quality. I am interested to see the advances in this.
patricklorio 1 year ago
wow I thought getting meshes was an open problem in opencv !
tarabaig 1 year ago
great work!
darksid3hack0r 2 years ago