Yu.A. Maniakov – Ph.D.(Eng.), Senior Research Scientist, 

Orel Branch of the FRC «Computer Science and Control» of RAS

E-mail: maniakov_yuri@mail.ru

O.A. Yakovlev – Junior Research Scientist, 

Orel Branch of the FRC «Computer Science and Control» of RAS E-mail: maucra@gmail.com

This paper presents an information model of incremental 3D reconstruction method. The method is based on implicit volumetric representation of the observed scene. It takes a live stream from stereo camera and fuses the frames into a single implicit surface model. The stream is processed frame by frame in an iterative manner. To process each frame we use accumulated data from previous iterations that forms a feedback loop.

Information model consists of processes such as: camera calibration; stereo matching; volumetric fusion; camera pose estimation; polygon mesh extraction.

Camera calibration is preliminary to 3D reconstruction and its results (reprojection matrix, rectification map, camera instrinsics) used as input data as well as stereo images. Stereo matching is a procedure that transforms stereo image into depth map. Volumetric fusion is a standard process of integrating new data into a volumetric model. Camera pose estimation is a procedure that takes current frame and volumetric model and computes camera extrinsics (position and orientation). Polygon mesh extraction is a process of obtaining explicit surface from an implicit volumetric model. Obtained surface is a reconstruction of the observed scene.

The quality of reconstruction depends on resolution of volumetric model (metric units per voxel) that is yet another input parameter for our model. This parameter allows us to improve quality in cost of increasing memory and time consumption and vice versa.

In addition, we implemented a 3D reconstruction software to illustrate correctness and practical applicability of the developed model.

1. Professional photogrammetry and drone-mapping Pix4D. URL = https://www.pix4d.com (data obrashcheniya: 08.11.2018).
2. ReconstructMe Real Time 3D Scanning Software. URL = http://reconstructme.net (data obrashcheniya: 08.11.2018).
3. Forsayt Devid A., Pons Zhan Komp’yuternoe zrenie. Sovremennyy podkhod: Per. s angl. M.: Izdatel’skiy dom «Vil’yams». 2004. 928 s. (In Russian).
4. Yakovlev O.A., Gasilov A.V. Uskorennyy algoritm stereosopostavleniya na osnove geodezicheskikh vspomogatel’nykh koeffitsientov. Informatika i ee primeneniya. 2016. T. 10. № 3. S. 98−104. (In Russian).
5. Curless B., Levoy M., A volumetric method for building complex models from range images. Proc. of the 23rd annual conference on Computer graphics and interactive techniques (SIGGRAPH '96). P. 303−312.
6. Nießner M., Zollhöfer M., Izadi S., Stamminger M. Real-time 3D reconstruction at scale using voxel hashing. ACM Trans. Graph. 2013.
   1. 32. № 6. Article № 169.
7. Newcombe R.A., Izadi S., Hilliges O., Molyneaux D., Kim D., Davison A.J., Kohli P., Shotton J., Hodges S., Fitzgibbon A. KinectFusion: Real-time dense surface mapping and tracking. Proc. of 10th IEEE International Symposium on Mixed and Augmented Reality (ISMAR '11). Washington, DC, USA: IEEE. 2011. P. 127−136.
8. Frisken S.F., Perry R.N. Efficient estimation of 3D Euclidean distance fields from 2D range images. Proc. of Symposium on Volume Visualization and Graphics. IEEE. 2002. P. 81−88.
9. Rusinkiewicz S., Levoy M., Efficient variants of the ICP algorithm. Proc. of the Third International Conference on 3-D Digital Imaging and Modeling. IEEE. 2001. P. 145−152.
10. Lorensen W.E., Harvey C.E. Marching cubes: A high resolution 3D surface construction algorithm. SIGGRAPH Comput. Graph. 1987.
    1. 21. № 4. P. 163−169.
11. Yakovlev O.A., Gasilov A.V. Sozdanie realistichnykh naborov dannykh dlya algoritmov trekhmernoy rekonstruktsii s pomoshch’yu virtual’noy s’‘emki komp’yuternoy modeli. Sistemy i sredstva informatiki. 2016. T. 26. № 2. S. 98−107. (In Russian).