Automatic extraction of road intersection points from USGS historical map series using deep convolutional neural networks

ABSTRACT

Road intersection data have been used across a range of geospatial analyses. However, many datasets dating from before the advent of GIS are only available as historical printed maps. To be analyzed by GIS software, they need to be scanned and transformed into a usable (vector-based) format. Because the number of scanned historical maps is voluminous, automated methods of digitization and transformation are needed. Frequently, these processes are based on computer vision algorithms. However, the key challenges to this are (1) the low conversion accuracy for low quality and visually complex maps, and (2) the selection of optimal parameters. In this paper, we used a region-based deep convolutional neural network-based framework (RCNN) for object detection, in order to automatically identify road intersections in historical maps of several cities in the United States of America. We found that the RCNN approach is more accurate than traditional computer vision algorithms for double-line cartographic representation of the roads, though its accuracy does not surpass all traditional methods used for single-line symbols. The results suggest that the number of errors in the outputs is sensitive to complexity and blurriness of the maps, and to the number of distinct red-green-blue (RGB) combinations within them.

KEYWORDS:

• Road network data
• object detection
• GIS
• Faster RCNN
• deep learning

Acknowledgments

The authors wish to extend their very special thanks to Dr. Kevin Raleigh from the University of Cincinnati and Dr. Jakub Nowosad from Adam Mickiewicz University for their valuable helps and comments which greatly improved the manuscript. We also appreciate the insightful comments from the associate editor, Shawn Laffan, and the anonymous reviewers.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data and codes availability statement

The data and codes that support the findings of this study are available in ‘https://doi.10.6084/m9.figshare.10282085.v1’. The code was captured from Tensorflow object detection API (https://github.com/tensorflow/models/tree/master/research/object detection). Also, the historical maps were collected from USGS topoView system (ngmdb.usgs.gov/topoview).

Additional information

Notes on contributors

Mahmoud Saeedimoghaddam

Mahmoud Saeedimoghaddam is a Ph.D. student at the Geography & GIS department of the University of Cincinnati. His research is focused on analyzing the spatial complex systems such as urban areas. He is also interested in using machine learning approach in addressing the spatial issues.

T. F. Stepinski

T. F. Stepinski is the Thomas Jefferson Chair Professor of Space Exploration at the University of Cincinnati and a Director of Space Informatics Lab. His recent area of research is a development of automated tools for intelligent and intuitive exploration of very large Earth and planetary datasets. He led the team who developed the GeoPAT2 a toolbox for pattern-based spatial analysis. He is also interested in computational approaches to geodemographics, racial segregation and diversity.