RESEARCH ON THE APPLICATION OF UAV OBLIQUE PHOTOGRAMMETRY TO LILONG HOUSING: TAKING MEILAN LANE AS AN EXAMPLE

Junyu Gao, Yuliang Shi, Yue Cai2023 · DOI

Abstract Vernacular architecture has always captured professionals’ attention for its unique approaches to construction and its integration of man and nature. Lilong houses (or Shikumen houses) are representative of Shanghai vernacular housing, combining traditional Chinese courtyard houses with Western townhouses. While UAV oblique photogrammetry has been applied to historic preservation in China, literature on Lilong houses remains scant. This research attempts to preserve and document Lilong houses using UAV-based 3D modeling, exploring a rational approach for narrow urban lanes. Using Meilan Lane as a case study, flight routes and angles were designed via AutoCAD and programmed into GoodStation. A total of 4,531 photos were processed to build four models at macro (40m–115m) and micro (20m–32m) levels. This paper elaborates on the pros and cons of UAV techniques for documenting Lilong residences and establishes an applicable approach for comprehensive heritage studies.

Introduction

Lilong housing represents a unique architectural fusion built primarily between 1876 and 1910. Today, these structures face dual threats: rapid urban decay and unauthorized modifications. Traditional documentation methods—manual measurement and on-site photography—are labor-intensive and often fail to capture the "top-down" perspective necessary for comprehensive preservation.

The physical environment of a Lilong complex presents significant challenges for standard aerial mapping. Lanes are extremely narrow, buildings are relatively tall (approx. 12.5m), and the airspace is cluttered with power lines, laundry poles, and thick tree canopies. This research seeks a "rational approach" to navigate these constraints and produce accurate digital twins for heritage conservation.

Spatial constraints of Meilan Lane, showing narrow sub-lanes ranging from 3.0m to 5.6m in width.

Methodology

The study employed an iterative, three-round experimental design to refine data collection using a DJI Mini 2 UAV, chosen for its small footprint and compliance with local flight regulations.

  1. Flight Planning: Initial routes were designed in AutoCAD to account for the "canyon" effect of the narrow lanes. Parameters such as speed, camera angles, and intervals were programmed into GoodStation to automate the drone’s path.
  2. Iterative Refinement: Initial rounds (30m to 118m) identified critical failures, such as image distortion on southern façades and obscurity caused by deep shadows.
  3. Optimized Workflow: The final phase utilized a multi-level approach: Macro-level flights at 40m and 115m for urban context, and Micro-level flights (20m–32m) specifically for the narrow lanes to capture architectural nuances.
Summary of Round 2 flight experiments, highlighting identified problems such as window distortion and shadow obscurity, along with proposed solutions.

Key Insights and Results

Overcoming Environmental Constraints

The researchers found that the height-to-width ratio of the lanes required a specialized 12-route flight design. This included "square tours" for roofs and specific flights along building footprints to capture façades while avoiding obstacles. Furthermore, the study concluded that cloudy days are superior to sunny days for photogrammetry in dense environments to avoid harsh shadows. A specific "seasonal window" (December to February) was identified as ideal, as local Platanus trees lose their leaves, making building envelopes fully visible.

Efficiency and Fidelity

The UAV workflow proved significantly more efficient than traditional methods, requiring only two people and four days to generate a complete model suite—a task that would take five people roughly two weeks using terrestrial reconnaissance. The resulting micro-level models provided enough resolution to assist craftsmen in locating degraded bricks, rotting window frames, and damaged gutters. Notably, the models successfully documented unauthorized third-floor additions and makeshift roof structures often hidden from ground-level inspection.

Conclusion and Future Work

This research demonstrates that a dual-height UAV approach is necessary to satisfy both urban planners and conservation architects. While the reliance on specific weather and seasonal windows presents a limitation, the efficiency gains are undeniable. Future work will focus on establishing a "data bank" of Lilong houses to support lifecycle management, tracking how these historic structures evolve or disappear during Shanghai's ongoing urban regeneration.

Summary updated

Abstract

Vernacular architecture has always captured professionals’ attention for its unique approaches to construction, its typical way of integrating man and nature and the way it is used. Being a combination of traditional courtyard houses in China with townhouses in the West, Lilong houses or Shikumen houses are representatives of Shanghai vernacular housing. In terms of applying UAV oblique photogrammetry to preserving and documenting historic architectures, some studies have been done in China, but scant literatures are found on Lilong houses. This research is an attempt to preserve and document Lilong houses by adopting UAV in 3D modelling, to explore a rational approach for the application of UAV oblique photogrammetry to narrow lanes and to establish a collection of models to document the ongoing changes as well as nuances of each building more carefully. In this research, Meilan Lane is chosen as an example. Aided by AutoCAD drawings, flight routes and angles are designed. parameters for each flight are pre-set in GoodStation to send quantified instructions to the drone. A total of 4,531 photos are gathered and 4 models are built by ContextCapture Master. Two models with flight height at 40m and 115m are generated for macro-level study and two models with flight height from 20m to 32m are for micro-level study. Through this research, this paper elaborates the pros and cons of using UAV techniques in documenting Lilong residences and tries to find an applicable approach to building other Lilong complex models for more comprehensive studies in the long run.
BibTeX
@inproceedings{gao2023research,
  author = {Junyu Gao and Yuliang Shi and Yue Cai},
  title = {{RESEARCH ON THE APPLICATION OF UAV OBLIQUE PHOTOGRAMMETRY TO LILONG HOUSING: TAKING MEILAN LANE AS AN EXAMPLE}},
  booktitle = {˜The},
  year = {2023},
  doi = {10.5194/isprs-archives-xlviii-m-2-2023-629-2023},
}
Figure