Abstract

This paper presents an approach for 3D reconstruction of heritage scenes using a videogrammetric-based device. The system, based on two video cameras with different characteristics, uses a combination of visual simultaneous localization and mapping (SLAM) and photogrammetry technologies. VSLAM, together with a series of filtering algorithms, is used for the optimal selection of images and to guarantee that the user does not lose tracking during data acquisition in real time. The different photogrammetrically adapted tools in this device and for this type of handheld capture are explained. An evaluation of the device is carried out, including comparisons with the Faro Focus X 330 laser scanner, through three case studies in which multiple aspects are analyzed. We demonstrate that the proposed videogrammetric system is 17 times faster in capturing data than the laser scanner and that the post-processing of the system is fully automatic, but takes more time than the laser scanner in post-processing. It can also be seen that the accuracies of both systems and the generated textures are very similar. Our evaluation demonstrates the possibilities of considering the proposed system as a new professional-quality measurement instrument.

Highlights

  • The wide-spread use of 3D models in fields such as architecture, urban planning, and archaeology [1,2,3] means that some professionals cannot generate 3D models without advanced technical knowledge in surveying

  • Among the various systems that are commonly used for data acquisition in architecture, urban planning, and archaeology, mid-range terrestrial laser scanners (TLS) [10] are commonly used, which are survey instruments that use laser pulses to measure distances and a rotating mirror to measure angles, generating a point cloud of the environment with a range of 0.5–300 m and root mean square error (RMS) between 1 and 4 mm

  • This scheme is based on simultaneous video capture by two cameras, A and B: With the images obtained with camera A, the camera pose is computed in real time and, the camera trajectory data are obtained

Read more

Summary

Introduction

The wide-spread use of 3D models in fields such as architecture, urban planning, and archaeology [1,2,3] means that some professionals cannot generate 3D models without advanced technical knowledge in surveying. The most widely used MMS for large-scale scanning and mapping projects in urban, architectural, and archaeological environments are vehicle-based mobile mapping platforms or those that must be transported by backpack or trolley In the former, we mention the UltraCam Mustang, manufactured by Vexcel Imaging GmbH (Gra, Austria); the Leica Pegasus:Two Ultimate, manufactured by Leica geosystems (Heerbrug, Switzerland); and the Trimble MX9, manufactured by Trimble (Sunnyvale, CA, USA) [23,24,25], which use inertial systems and GNSS for positioning, and are equipped with LIDAR profilers with high scanning speeds (up to 500 scans/s) and 360 cameras to achieve street-level scenery with geo-positioned panoramic imagery.

Materials and Methods
Cameras Calibration
Data Acquisition and Preliminary Images Selection
Image Selection by Filtering Process
Photogrammetric Process
Point Cloud Filtering
Mesh and Texture Mapping Generation
Experimental Test
Comparison Times for Data Acquisition and Processing
Points Cloud Resolution and Distribution
Accuracy Assessments
Control Points Accuracy Test
Analysis of Systematic Errors Using Circular Statistics
Cross-Sections
Points Color Evaluation
Conclusions
Findings
24. Leica Pegasus
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call