iv-image.com

Research at iv-image.com

Currently I am working on 3D volume recognition. That is to extract 3D data from 2D images and reconstruct 3D models with texture mapping, not for 3D vision only. This technology can be used at 3D Scanner or similar products. The advantages of my method are:

The algorithm is very stable and reliable without any wrong 3D information.
Does not need any other assisting method, such as Laser scanning.
All 3D data extracted from photos only and the photos also used for texture mapping.

The followings are a part of samples for extracted 3D models (The red line in figures shows x-axis, the green y-axis, the blue z-axis).


The images show a 3D model in different directions for an object with holes . This model was extracted from 3 photos.


This is another example. The left 2 images show the extracted 3D model in different directions. The model has totally 23,785,390 points that is maximum resolution based on the photos (1024 x 768 pixels). The resolution of 3D model can be controlled by users based on the resolution of camera, similar to the DPI value on flatbed scanners. These points were extracted from four photos like the right image taken by a digital camera.

This image shows the wireframe of a 3D model extracted from four photos. The more images used, the more accuracy of 3D shapes can be obtained. The 3D object will be saved in the Wavefront OBJ format (.obj). The OBJviewer on the software page can be used to display the 3D objects. The more 3D formats could be used in future.