From Data to Information: Evaluating Attribute-Enriched Point Clouds for Accurate Urban Corner and Edge Measurement
Yung-Ching Yang(a*), Yu-Yun Chen(b), Jen-Jer Jaw(a)

a) Department of Civil Engineering, National Taiwan University
No. 1, Sec. 4, Roosevelt Rd., Taipei 106319, Taiwan
*ivy.ycyang[at]gmail.com
b) Chinese Society of Photogrammetry & Remote Sensing, CSPRS
3F., No. 111, Sec. 5, Roosevelt Rd., Taipei 11681 , Taiwan

Abstract

Accurate measurement of building corners and edges from point clouds is essential for advancing automation in urban mapping. Traditional point clouds-composed purely of 3D geometric coordinates-often lack the contextual cues to identify structural features in dense urban scenes.
This study investigates transforming point cloud data into information by integrating attribute information during the point cloud generation and fusion process. Emphasizing the preservation and propagation of attribute information, starting from the multi-view image domain, through the point cloud generation process, and ultimately into the final 3D representation.
We extract attribute information directly from the source imagery, such as corner-like features and semantic labels. These features are retained and embedded as attributes throughout the multi-view fusion process, resulting in attribute-enriched point clouds that combine spatial geometry with descriptive metadata.
This study compares two point cloud generation modes: (1) geometry-only point clouds, and (2) enriched point clouds containing feature traces and scene understanding. The enriched data allows for spatial measurement and provides higher-level interpretability, improving the clarity and confidence of corner and edge identification. Instead of relying on purely geometric saliency at the end stage, we analyze how information preserved from the image domain can influence feature selection within the 3D data.
Furthermore, we demonstrate that these enriched point clouds can be organized into structured attribute tables, facilitating the automation of feature selection and information retrieval in future workflows. Experimental results show that attribute-integrated point clouds exhibit greater reliability in identifying urban structural features. These findings underline the shift from passive 3D capture toward active, information-driven point cloud modeling that bridges the gap between perception and measurement.

Keywords: Attribute-Enriched Point Cloud- Multi-View Image Fusion- Corner and Edge Measurement- Semantic-Guided Reconstruction- 3D Spatial Information Modeling

Topic: Topic B: Applications of Remote Sensing