|
Presenting a model for generalizing and simplifying road networks using graph-based deep learning methods
|
Marzie Zarei   , Mohammad Karimi * , Parastoo Pilehforooshha |
|
|
|
Abstract: (10 Views) |
One of the core responsibilities of national mapping organizations is the production of maps at multiple scales, a task that is inherently time-consuming and costly. An effective approach to reducing both time and expense is the generation of small-scale maps through the generalization and simplification of large-scale maps. In this research, considering the crucial role of road networks in topographic databases, a three-step framework was developed for modeling the generalization and simplification of road networks. In the first step, to address the limitations of existing methods, a hybrid structure was proposed that integrates the Mesh, Stroke, and building-block approaches. In this stage, the road network was decomposed into linear and surface-linear structures, and their topological relationships were extracted using a dual graph. In the second step, thematic, geometric, and topological features of the structures were extracted and standardized, after which classification was performed using the Random Forest algorithm for linear structures and the TAGCN method for surface-linear structures. Based on the classification results, less significant or isolated roads were eliminated, and the remaining lines were simplified and smoothed according to the target scale. In the third step, to finalize the generalization and simplification process, building blocks were generalized to the target scale, and spatial conflicts between them and the resulting road network were detected and resolved. The results demonstrated that the overall classification accuracy was 88% for linear structures and 93% for surface-linear structures. Furthermore, the evaluation revealed that major alterations compared to the original network were limited to approximately 8% of the network’s length and density. |
Article number: 5 |
| Keywords: Generalization and Simplification, Road Networks, Topographic Database, Graph-based Deep Learning, Elimination–Selection Operator |
|
|
|
Type of Study: Research |
Subject:
GIS
Received: 2025/10/5 | Accepted: 2026/02/7
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
