Bamboo structure monitoring with RPAS and artificial vision

An article that recently appeared in Advances in bamboo science proposed an innovative method using Remotely Piloted Aircraft Systems (RPAS) and artificial vision to monitor a bamboo structure and identify the critical issues present. It was part of a joint FISR (Fiscal Implications of Structural Reforms) project between China and Italy.

Background

Bamboo has recently received much attention as a building material due to its unique properties such as fast and easy cultivation, flexibility, resistance, lightweight and durability. However, its widespread application in construction is limited due to sustainability challenges as it is a natural material. It is sensitive to attacks from insects and animals, and from weather influences such as rain and sun.

Thus, tailor-made treatments, careful precautions and regular structural inspections are crucial to immediately detect any deterioration and intervene accordingly, ensuring the longevity of bamboo structures. Artificial intelligence and machine learning are increasingly used in structural health monitoring to reveal variations or anomalies and identify potential problems. These algorithms can also predict structural deterioration using historical data and other parameters related to the environment and structure under investigation.

Moreover, sharing information about bamboo construction with the widest possible audience (infrastructure management personnel and simple users who want to know its potential) is crucial to promote its use in sustainable construction. Special virtual/augmented/mixed reality applications (VR/AR/MR app) can be used to digitize and disseminate such information in real time. Thus, this article focuses on the real-time information dissemination together with structural health monitoring of bamboo structures.

Methods

This study was conducted on a bamboo roof structure at the entrance of Zhejiang University, China. The process involves a series of steps: an overview of the structure, 3D modeling based on the images of the structure, and importing the model and other structural information into a VR/AR/MR app for analysis.

First, the structure was surveyed using a DJI Mavic 2 Pro drone, which uses a high-resolution Global Positioning System (GPS) integrated camera to geotag each photo and record its location information. The drone images were then processed with commercial Agisoft Metashape software based on the Structure for Motion algorithm. This process allows the reconstruction of a point cloud using photos taken by drones, from which a detailed and accurate 3D model can be created.

A system was then implemented that used the you look only once (YOLO) v5s6 algorithm, based on a deep learning object detection model, to monitor the bamboo structure. It is an open-source customizable algorithm with high speed and accuracy in real-time object detection. Additionally, YOLO’s easy integration with drones helps capture structural deterioration and identify issues in the bolted joints.

Finally, a VR/AR/MR app developed by the authors was used to analyze different parts of the bamboo structure. The app uses the Microsoft HoloLens tool to display the hologram of the research phases of the structure. It improves the performance of structure monitoring and management systems through in-depth analysis and close-up views of the components without the need for their physical presence. In addition, the app helps spread information about the possible use of bamboo in construction technology.

Results

The 30-minute drone survey captured 459 photos, which were converted into a 3D model by Agisoft Metashape software in a further 7.2 hours. After importing the 3D model into the VR/AR/MR app, the metal connections of the structure under investigation were analyzed and categorized as degraded (slips and fractures) or non-deteriorated. A technical sheet of the materials or structural building information modeling (BIM) can be obtained through the app and shared with multiple users at the same time.

The information disseminated through the VR/AR/MR app can help maintenance personnel and private individuals understand the properties of bamboo structures. This application provides operators with an immersive and interactive method to explore details and specific areas of construction without a physical presence on site, reducing travel requirements and exposure to potentially hazardous or difficult-to-reach areas.

Additional information on the state of degradation is provided by the VR/AR/MR app by integrating the YOLO data, increasing accuracy while assessing structural conditions. It facilitates the timely planning of maintenance and intervention activities. Such real-time monitoring of the structure can help identify loose bolts and ensure the safety and durability of bamboo structures.

Conclusion

Overall, the methodologies tested and proposed in this study performed well in the investigation and evaluation of bamboo roof structure in China. An automated deep learning-based model has been developed to capture and identify degradations using bounding boxes. The long processing times in this study reflect the complexity of the structure and the richness of the results recorded. The authors proposed parallel computation where the workload was distributed across multiple processors to minimize processing time.

The Geomatics approach of this study is important for monitoring bamboo structures and can efficiently help solve problems related to the possible structural vulnerability and limitations in bamboo application areas. It can better highlight the potential of bamboo as an environmentally friendly building material and promote its use in environmentally friendly construction practices.

Journal reference

Barrile, V., and Genovese, E. (2024). Bamboo structures: innovative methods and applications for structural health monitoring and dissemination. Advances in bamboo science, 100079–100079. https://doi.org/10.1016/j.bamboo.2024.100079, https://www.sciencedirect.com/science/article/pii/S2773139124000247