THESIS
2023
1 online resource (xii, 64 pages) : illustrations (some color)
Abstract
Point cloud is becoming increasingly important in various fields, including robotics,
computer vision, and 3D reconstruction. It is also used in the construction industry for
surveying and quality control purposes. To acquire a dense and high-quality point cloud, the
most common way is to use a LiDAR scanner. However, the non-colorized raw points from
sensor is very difficult for users to distinguish in visualization. Point cloud colorization
refers to the process of assigning points with the physical color in the real world. High
quality colorized point cloud can achieve photo-realistic rendering effects that supports
various application such as digital twins, 3D modeling and virtual reality.
In this paper, we introduce two manners of point cloud colorization, online and offline.
We il...[
Read more ]
Point cloud is becoming increasingly important in various fields, including robotics,
computer vision, and 3D reconstruction. It is also used in the construction industry for
surveying and quality control purposes. To acquire a dense and high-quality point cloud, the
most common way is to use a LiDAR scanner. However, the non-colorized raw points from
sensor is very difficult for users to distinguish in visualization. Point cloud colorization
refers to the process of assigning points with the physical color in the real world. High
quality colorized point cloud can achieve photo-realistic rendering effects that supports
various application such as digital twins, 3D modeling and virtual reality.
In this paper, we introduce two manners of point cloud colorization, online and offline.
We illustrate our implementation of the two manners with different principles. By specially
tailoring necessary components, our online method achieves the goal of real-time and
memory-efficiency, providing a fast and robust preview that significantly helps scanners improve
the completion of scanning. Our offline method contains several improvements and
novel approaches including image selection mechanism, a novel visibility check algorithm, and evaluation based merge policy. Experiments show that our novel visibility check algorithm
is much faster and more robust to dynamic points than traditional method. Our
whole offline colorization pipeline outperforms several commercial products and achieve
comparable quality as the state of art commercial point cloud processing solution.
Post a Comment