A novel binocular vision system for accurate 3-D reconstruction in large-scale scene based on improved calibration and stereo matching methods.

Autor: Wang, Daolei, Sun, Hao, Lu, Wu, Zhao, Wenbin, Liu, Yiteng, Chai, Pingping, Han, Yang
Předmět:
Zdroj: Multimedia Tools & Applications; Jul2022, Vol. 81 Issue 18, p26265-26281, 17p
Abstrakt: The 3D reconstruction technology based on stereo vision directly acquires the 3D model of the object through two 2D images. The reconstruction is recovered highly automated. It does not require any prior information and special hardware support. However, for large outdoor scenes, the existing 3D reconstruction technology based on stereo vision often has detailed information loss and data scattering, which makes the reconstruction result less accurate. For this problem, a novel binocular vision system for 3D reconstruction in large-scale scene is proposed. This system uses the calibration rods to perform the calibration calculation based on the polar line correction, and then it utilizes the weighted least squares filter(WLSF) to denoise and smooth the depth map, finally, the point cloud is reconstructed. The results of experiment show that compared with the traditional stereo vision system, the calibration results of new system is more accurate and the calibration space is expanded. The depth map is smoother and less noisy. The system can reconstruct the 3D point cloud of large scene more stably and accurately, has high practical value. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: