The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Download
Citation
Articles | Volume XLI-B3
https://doi.org/10.5194/isprs-archives-XLI-B3-481-2016
https://doi.org/10.5194/isprs-archives-XLI-B3-481-2016
09 Jun 2016
 | 09 Jun 2016

CONTINUOUS MAPPING OF TUNNEL WALLS IN A GNSS-DENIED ENVIRONMENT

Michael A. Chapman, Cao Min, and Deijin Zhang

Keywords: Tunnel mapping, image mosaicking, image bridging, visual odometry, GNSS-denied

Abstract. The need for reliable systems for capturing precise detail in tunnels has increased as the number of tunnels (e.g., for cars and trucks, trains, subways, mining and other infrastructure) has increased and the age of these structures and, subsequent, deterioration has introduced structural degradations and eventual failures. Due to the hostile environments encountered in tunnels, mobile mapping systems are plagued with various problems such as loss of GNSS signals, drift of inertial measurements systems, low lighting conditions, dust and poor surface textures for feature identification and extraction. A tunnel mapping system using alternate sensors and algorithms that can deliver precise coordinates and feature attributes from surfaces along the entire tunnel path is presented. This system employs image bridging or visual odometry to estimate precise sensor positions and orientations. The fundamental concept is the use of image sequences to geometrically extend the control information in the absence of absolute positioning data sources. This is a non-trivial problem due to changes in scale, perceived resolution, image contrast and lack of salient features. The sensors employed include forward-looking high resolution digital frame cameras coupled with auxiliary light sources. In addition, a high frequency lidar system and a thermal imager are included to offer three dimensional point clouds of the tunnel walls along with thermal images for moisture detection. The mobile mapping system is equipped with an array of 16 cameras and light sources to capture the tunnel walls. Continuous images are produced using a semi-automated mosaicking process. Results of preliminary experimentation are presented to demonstrate the effectiveness of the system for the generation of seamless precise tunnel maps.