Abstract:Historical maps are invaluable for analyzing long-term changes in transportation and spatial development, offering a rich source of data for evolutionary studies. However, digitizing and classifying road networks from these maps is often expensive and time-consuming, limiting their widespread use. Recent advancements in deep learning have made automatic road extraction from historical maps feasible, yet these methods typically require large amounts of labeled training data. To address this challenge, we introduce a novel framework that integrates deep learning with geoinformation, computer-based painting, and image processing methodologies. This framework enables the extraction and classification of roads from historical maps using only road geometries without needing road class labels for training. The process begins with training of a binary segmentation model to extract road geometries, followed by morphological operations, skeletonization, vectorization, and filtering algorithms. Synthetic training data is then generated by a painting function that artificially re-paints road segments using predefined symbology for road classes. Using this synthetic data, a deep ensemble is trained to generate pixel-wise probabilities for road classes to mitigate distribution shift. These predictions are then discretized along the extracted road geometries. Subsequently, further processing is employed to classify entire roads, enabling the identification of potential changes in road classes and resulting in a labeled road class dataset. Our method achieved completeness and correctness scores of over 94% and 92%, respectively, for road class 2, the most prevalent class in the two Siegfried Map sheets from Switzerland used for testing. This research offers a powerful tool for urban planning and transportation decision-making by efficiently extracting and classifying roads from historical maps.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the challenges faced in extracting and classifying road networks from historical maps. Specifically: 1. **High cost and time - consuming**: - Digitizing and classifying road networks in historical maps usually require a large amount of manual annotation work, which is not only costly but also time - consuming. Therefore, it limits the wide application of these maps. 2. **Lack of large - scale labeled data**: - Current methods usually require a large amount of labeled training data to achieve automatic road extraction, but the labeled data of historical maps is very limited and difficult to obtain. 3. **Distribution shift problem**: - When using synthetic data for training, the distribution difference between synthetic data and real data may lead to poor performance of the model in practical applications. Therefore, how to effectively deal with this distribution shift is an important problem. ### Solutions To address the above challenges, the author proposes a new framework that combines deep learning, geoinformatics, computer graphics and image processing methods to achieve the function of extracting and classifying roads from historical maps. The specific steps are as follows: 1. **Cascade training of binary segmentation models**: - First, use the cascade training method to train a binary segmentation model to extract road geometries from historical maps. This process includes pre - training the model on a larger dataset and then fine - tuning it on historical map data. 2. **Morphological operations, skeletonization, vectorization and filtering**: - After extracting the road geometries, perform morphological operations, skeletonization, vectorization and filtering to optimize the road extraction results. 3. **Synthetic training data generation**: - Through the computer graphics function, manually redraw road segments and use predefined road category symbols to generate synthetic training data. 4. **Deep ensemble model training**: - Use synthetic data to train a deep ensemble model to generate class probabilities for each pixel to alleviate the distribution shift problem. 5. **Road classification**: - Discretize the predicted probabilities along the extracted road geometries and further process them to classify the entire road, identify potential road category changes, and finally generate a labeled road category dataset. ### Experimental results This method was tested on the Siegfried map in Switzerland. For the most common road category 2, it achieved more than 94% completeness and 92% accuracy. The research results show that this method can efficiently extract and classify roads from historical maps and has broad application prospects, especially in urban planning and traffic decision - making.

Probabilistic road classification in historical maps using synthetic data and deep learning

Towards the automated large-scale reconstruction of past road networks from historical maps

Novel Framework for 3D Road Extraction Based on Airborne LiDAR and High-Resolution Remote Sensing Imagery

Classification of urban morphology with deep learning: Application on urban vitality

A Layered Approach for More Robust Generation of Road Network Maps from Vehicle Tracking Data

Potential of deep learning segmentation for the extraction of archaeological features from historical map series

Automatic Extraction of Relevant Road Infrastructure using Connected vehicle data and Deep Learning Model

Detecting Hardly Visible Roads in Low-Resolution Satellite Time Series Data

Image-based machine learning and cluster analysis for urban road network: employing Orange for codeless visual programming

Deep learning approaches for delineating wetlands on historical topographic maps

A deep learning framework for road marking extraction, classification and completion from mobile laser scanning point clouds

A Survey of Deep Learning Road Extraction Algorithms Using High-Resolution Remote Sensing Images

Automatic Road Marking Extraction and Vectorization from Vehicle-Borne Laser Scanning Data

Road Mapping in Low Data Environments with OpenStreetMap

Combining Deep Learning and Mathematical Morphology for Historical Map Segmentation

Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection

Large-scale Weakly Supervised Learning for Road Extraction from Satellite Imagery

Improving classification of road surface conditions via road area extraction and contrastive learning

DeepAerialMapper: Deep Learning-based Semi-automatic HD Map Creation for Highly Automated Vehicles

Fusion of neural networks, for LIDAR-based evidential road mapping

TPMT based Automatic Road Extraction from 3D Real Scenes