A Novel Framework for Road Data Surveying Based on Deep-learning and GIS

This paper proposes a novel framework for road data surveying and 3D urban visualisation based on deep-learning technologies, open data, and GIS techniques. Existing road inventory preparation methods are time-consuming, labour-intensive, inefficient, and lack an acceptable method for 3D urban visualisation in Sri Lanka. Thus, modern approaches are not applicable due to a lack of resources, technology, and financial capacity; the proposed framework will enable us to overcome these weaknesses.

The study comprised of three main stages: a literature review, development of the framework, and its validation in Ranna area. The framework was applied to two consecutive model validation events: the first related to the road data surveying model and the second to the 3D urban visualisation model. Its proposed returned KAPPA Accuracy Scores were 92% and 90% respectively.

The findings of this study further the use of cutting-edge deep learning and mapping techniques in transport and urban planning, making the preparation of road inventory surveys and 3D urban visualisations more cost-effective and efficient. Transport engineers, urban and regional planners, geographers, and GIS experts can employ the proposed framework for road data collection and 3D urban visualisation.