Abstract:As the number of vehicles and the volume of traffic swell in urban centers, cities have experienced a concomitant increase in traffic accidents. Proactively identifying accident-prone hotspots in urban environments holds the promise of preventing traffic mishaps, thereby curtailing the incidence of accidents and reducing property damage. This research introduces the Two-Branch Contextual Feature-Guided Converged Network (TCFGC-Net) utilizing multimodal satellite and street view data. Designed to extract global structural features from satellite imagery and dynamic continuous features from street view imagery, the model aims to improve the accuracy of detecting urban accident hotspots. For the satellite imagery branch, we propose the Contextual Feature Coupled Convolutional Neural Network (Trans-CFCCNN) designed to extract global spatial features and discern feature correlations across adjacent regions. For the street view imagery branch, we develop the Sequential Feature Recurrent Attention Network (SFRAN) to assimilate and integrate dynamic scene features captured from successive street view images. We designed the Multi-Branch Feature Adaptive Fusion Structure (MBFAF) to aggregate different branch features for accurate identification of accident hotspots. Experimental results show that the model performs well, with an overall accuracy of 93.7 %. Ablation studies confirm that relative to standalone street view and satellite branch analyses, implementing multimodal fusion enhances the model's accuracy by 12.05 % and 17.86 %, respectively. The innovative fusion structure proposed herein garners a 4.22 % increase in model accuracy, outpacing conventional feature concatenation techniques. Furthermore, the model outperforms existing deep learning models in terms of overall efficacy. Additionally, to showcase the efficacy of the proposed model structure, we utilize Class Activation Maps (CAM) to provide visual interpretability for the model. These results suggest that the dual-branch fusion model effectively decreases false alarm occurrences and directs the model's focus toward regions more pertinent to accident hotspots. Finally, the code and model used for identifying hotspots of urban traffic accidents in this study are available for access: https://github.com/gwt-ZJU/TCFGC-Net.

Estimation of Road Accident Severity Using K-Means Clustering of Spatio-Temporal Data with Backend Spatial Data Infrastructure

Spatiotemporal Clustering and Analysis of Road Accident Hotspots by Exploiting GIS Technology and Kernel Density Estimation

Modeling Severe Traffic Accidents With Spatial And Temporal Features

Hierarchical clustering analysis framework of mutually exclusive crash causation parameters for regional road safety strategies

Factors propelling fatalities during road crashes: A detailed investigation and modelling of historical crash data with field studies

Analysis of Spatiotemporal Factors Affecting Traffic Safety Based on Multisource Data Fusion

Identifying accident prone areas and factors influencing the severity of crashes using machine learning and spatial analyses

Discovering Spatio-Temporal Clusters of Road Collisions Using the Method of Fast Bayesian Model-Based Cluster Detection

Evaluation of the Impact of Spatial and Environmental Accident Factors on Severity Patterns of Road Segments

Phase-wise injury integrated severity modeling of road accidents: a two-stage hybrid multi-criteria decision-making model

Investigating the fatal pedestrian crash occurrence in urban setup in a developing country using multiple-risk source model

A two-phase clustering approach for traffic accident black spots identification: integrated GIS-based processing and HDBSCAN model

Fusion of Satellite and Street View Data for Urban Traffic Accident Hotspot Identification

Spatial multiresolution analysis approach to identify crash hotspots and estimate crash risk

SENE: A novel manifold learning approach for distracted driving analysis with spatio-temporal and driver praxeological features

An Adaptive Clustering Approach for Accident Prediction

Identification and analysis of accident black spots using Geographic Information System (GIS): A study on Kushtia-Jhenaidah national highway (N704), Bangladesh

Exploring the severity of bicycle-vehicle crashes using latent class clustering approach in India

A fatal point concept and a low-sensitivity quantitative measure for traffic safety analytics

K-Means Algorithm for Clustering The Location Of Accident-Prone On The Highway

Identification of Traffic Accident Hotspots using Geographical Information System (GIS)