Identifying safe intersection design through unsupervised feature extraction from satellite imagery

  title={Identifying safe intersection design through unsupervised feature extraction from satellite imagery},
  author={Jasper S. Wijnands and Haifeng Zhao and Kerry A. Nice and Jason Thompson and Katherine Scully and Jingqiu Guo and Mark R. Stevenson},
  journal={Computer‐Aided Civil and Infrastructure Engineering},
  pages={346 - 361}
The World Health Organization has listed the design of safer intersections as a key intervention to reduce global road trauma. This article presents the first study to systematically analyze the design of all intersections in a large country, based on aerial imagery and deep learning. Approximately 900,000 satellite images were downloaded for all intersections in Australia and customized computer vision techniques emphasized the road infrastructure. A deep autoencoder extracted high‐level… 

Sensing accident-prone features in urban scenes for proactive driving and accident prevention

A visual notification of accident-prone features to drivers, based on real-time images obtained via dashcam, is proposed, using Google Street View images around accident hotspots identified by accident dataset to train a family of deep convolutional neural networks (CNNs).

Urban feature analysis from aerial remote sensing imagery using self-supervised and semi-supervised computer vision

Analysis of overhead imagery using computer vision is a problem that has re-ceived considerable attention in academic literature. Most techniques that operate in this space are both highly

Self-Supervision. Remote Sensing and Abstraction: Representation Learning Across 3 Million Locations

This work explores contrastive representation learning methods on the task of imagery-based city classification, an important problem in urban computing, and shows that self-supervised methods can build a generalizable representation from as few as 200 cities, with representations achieving over 95% accuracy in unseen cities with minimal additional training.

Small Object Detection and Tracking in Satellite Videos With Motion Informed-CNN and GM-PHD Filter

The proposed track-by-detection approach to detect and track small moving targets by using a convolutional neural network and a Bayesian tracker outperforms competing approaches across different scenarios with both object detection and object tracking metrics.

Sources and Applications of Emerging Active Travel Data: A Review of the Literature

This paper reviews relevant research on AT outcomes to address the potential of AT and emerging data sources in supporting the transport paradigm shift toward AT, and assess the emergingData sources in terms of their applications and potential limitations.

Assessing Spatial Variations of Traffic Congestion Using Traffic Index Data in a Developing City: Lessons from Johannesburg, South Africa

The COVID-19 pandemic has created unforeseen effects in public transport and the mobility of people in cities globally. Johannesburg, being a developing city in one of the most affected countries in

A self-adaptive multi-objective feature selection approach for classification problems

A self-adaptive multi-objective genetic algorithm (SaMOGA) is proposed, which is designed to maintain a high performance even when the dimensionality of the datasets grows, and can guarantee classification performance while removing many features.



Combining Satellite Imagery and Open Data to Map Road Safety

A deep learning-based mapping approach that leverages open data to learn from raw satellite imagery robust deep models able to predict accurate city-scale road safety maps at an affordable cost is proposed.

Street smarts: measuring intercity road quality using deep learning on satellite imagery

A model for monitoring the quality of road infrastructure using satellite imagery is developed, enabling much larger scale and much lower costs than are achievable with current methods, and providing insight into the contribution of high-quality roads to many societal development indicators.

Unsupervised Deep Learning to Explore Streetscape Factors Associated with Urban Cyclist Safety

This research uses publicly available big data such as maps and satellite images to capture information of the environment of cyclist crashes and suggests a deep learning method is able to learn the hidden features from map andatellite images and model the road environments using GANs.

Road Network Extraction and Intersection Detection From Aerial Images by Tracking Road Footprints

An automatic road seeding method based on rectangular approximations to road footprints and a toe-finding algorithm to classify footprints for growing a road tree and introduces a lognormal distribution to characterize the conditional probability of A/P ratios of the footprints in the road tree.

Road network extraction: a neural-dynamic framework based on deep learning and a finite state machine

A neural-dynamic tracking framework is proposed to extract road networks based on deep convolutional neural networks (DNN) and a finite state machine (FSM) and the results illustrate that this approach is more accurate and efficient than the traditional ones.

Use of Satellite Imagery for Establishing Road Horizontal Alignments

A new method for establishing road horizontal alignment using IKONOS 1 m spatial resolution imagery is presented, which performs the search procedures in a much smaller area than the image size and achieves faster computations.

Real-time monitoring of driver drowsiness on mobile platforms using 3D neural networks

It is shown how depthwise separable 3D convolutions, combined with an early fusion of spatial and temporal information, can achieve a balance between high prediction accuracy and real-time inference requirements.

Recent progress in road and lane detection: a survey

This paper presents a generic break down of the problem of road or lane perception into its functional building blocks and elaborate the wide range of proposed methods within this scheme.