Journal of Engineering, Project, and Production Management, 2026, 16(6), 2025-223
3D Extraction of Urban Heritage Elements Using Polarized Self-Attention and Enhanced Neural Radiance
1 Instructor,
School of Art and Design, Wuhan Institute of Technology, Wuhan, 430205,
China, E-mail: xiaofeichen0666@126.com (corresponding author).
2 Professor, School of Art and
Design, Wuhan Institute of Technology, Wuhan, 430205, China
Project Management
Received October 3, 2025; revised December 28, 2025; accepted June 9, 2026
Available online June 21, 2026
Abstract: With the growing need for precise extraction and digital documentation of urban historical landscape elements, such as ancient building brackets, city wall textures, traditional street layouts, and stone inscription patterns, this study proposes a novel extraction algorithm based on improved polarized self-attention and enhanced neural radiance fields. It tackles technical challenges by jointly optimizing 3D scene representation and fine-grained feature extraction, enabling accurate and robust element identification in complex environments. Tests on real-world urban historical landscapes show that the model outperforms existing methods. For 3D digital documentation of ancient building brackets, it achieves an average error of 0.9 to 1.8mm, detail completeness of 93.8% to 99.8%, and texture clarity above level 4.2. In wall texture recognition, accuracy exceeds 96.9%, with lesion localization errors of 1.2 to 2.8cm. The method significantly improves extraction precision and anti-interference capability. It not only extracts elements accurately but also generates 3D semantic models, offering core technical support for historical landscape restoration and urban heritage management. This provides a new technological pathway for digital preservation and sustainable urban development.
Keywords: Polarized self-attention mechanism, neural radiance fields, urban historical landscape element extraction, three-dimensional scene representation, multi-source data.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Chen, X., and Guo, L. (2026). 3D Extraction of Urban Heritage Elements Using Polarized Self-Attention and Enhanced Neural Radiance. Journal of Engineering, Project, and Production Management, 16(6), 2025-223.
