A Semi-Automated Technique for Cadastral Boundary Extraction from UAV Images Using Deep-Learning and Geospatial Techniques
No Thumbnail Available
Date
2024-06-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The 2030 Agenda for Sustainable Development acknowledges the crucial role of land in advancing and accomplishing the Sustainable Development Goals across the globe. Nevertheless, a large portion of land rights worldwide are still unregistered in government-sanctioned systems. To address this issue, the Fit-for-Purpose (FFP) approach to land administration has been introduced. This approach aims to streamline cadastral mapping and minimize the expenses and time associated with conventional surveying methods. This study examines the progress and possibilities of using Unmanned Aerial Vehicle (UAV) imagery and Deep learning techniques, particularly Convolutional Neural Networks (CNNs), which are employed for the extraction of cadastral boundaries. CNNs have demonstrated their effectiveness in accurately and efficiently extracting boundaries, as they are capable of extracting high-level features without the need for human expertise in feature engineering. The study tested the BDCN and HED deep learning models for cadastral boundary extraction from UAV datasets. The BDCN model achieved an average precision of 0.68, a recall of 0.80, and an F-score of 0.73. It had an average precision of 0.88 and an overall IoU of 0.85. The HED model performed slightly better achieving an average precision of 0.66, a recall of 0.68, and an F-score of 0.67. It also demonstrated an average precision of 0.98 and an overall Intersection over Union (IoU) of 0.88. The results indicate that these deep learning models can effectively extract cadastral boundaries in vector polygon format, which can be directly used in mapping for rural cadaster with post-processing and field verification. The study highlights the potential of using UAV imagery and deep learning techniques to support more efficient and cost-effective cadastral boundary mapping, aligning with the goals of the Fit-for-Purpose land administration approach.
Keywords: Land administration, Cadastral mapping, Unmanned Aerial Vehicle (UAV), Deep learning techniques
Description
Keywords
Land administration, Cadastral mapping, Unmanned Aerial Vehicle (UAV), Deep learning techniques