Open Access
Subscription or Fee Access
UAS-Based 3D Reconstruction Imagery Error Analysis
Abstract
With the rapid development of navigation, guidance and flight control systems, and sensing technologies, lately the market of small Aerial Systems (UASs) has boomed. UASs have been deployed in many different fields, including inspection of civil infrastructure conditions. To monitor structural performance, a UAS will be equipped with different sensing devices based on the task and desired data product. So far, the two most commonly used sensors on UAS are optical cameras and Light Detection and Ranging (LIDAR) systems. By processing images of structures captured from a hovering UAS, a better understanding of structural conditions such as structural defects sizes or locations, or structural displacement may be achieved. For residential or commercial buildings, people often use 3D models reconstructed from multiple 2D images to analyze and assess structural conditions. The 3D models obtained directly from 2D images do not have an absolute scale, and they are georeferenced via integration with ground control points or with additional sensors, such as LIDAR system. The quality of structure assessment is highly dependent on the accuracy of the 3D reconstruction and geo-referencing. However, various types of issues can arise in UAS-based 3D reconstruction, for instance, from the real-time image collection process, image and feature processing, model estimation, and integration with sensors or control points. In this paper, we will identify the major error sources, then quantify or simulate these errors, and estimate their contributions to the overall error in the structure assessment process. A set of imageries of regular shape models captured from the LIDAR onboard a small UAS will be collected and analyzed. These representative images will be used as examples to visualize and quantify the error sources of structural monitoring. Based on the error analysis results, we will discuss approaches to potentially constrain the error sources and mitigate their impact on the assessment process.
DOI
10.12783/shm2019/32248
10.12783/shm2019/32248