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.

A classification model of road accumulated snow severity based on the fusion of DeepLabV3 + and U-Net

Unifying Terrain Awareness Through Real-Time Semantic Segmentation

Road Surface Translation Under Snow-Covered and Semantic Segmentation for Snow Hazard Index

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

A system of vision sensor based deep neural networks for complex driving scene analysis in support of crash risk assessment and prevention

Automated Recognition of Snow-Covered and Icy Road Surfaces Based on T-Net of Mount Tianshan

AI monitoring and warning system for low visibility of freeways using variable weight combination model

Roadway Snow Detection Using Dual-Spectrum Camera Images and Computer Vision

An Adaptive Machine Learning Framework for Multi-Scenes Road Surface Weather Condition Monitoring

Deep Dense Multi-scale Network for Snow Removal Using Semantic and Geometric Priors

Winter Road Surface Condition Recognition Using A Pretrained Deep Convolutional Network

Design of Efficient Deep Learning models for Determining Road Surface Condition from Roadside Camera Images and Weather Data

Road Weather Condition Recognition via Fusing Images and Upsampled Point Cloud Reflection Intensities

DesnowNet: Context-Aware Deep Network for Snow Removal

Snowy Night-to-Day Translator and Semantic Segmentation Label Similarity for Snow Hazard Indicator

L-DIG: A GAN-Based Method for LiDAR Point Cloud Processing under Snow Driving Conditions

Snow‐ or ice‐covered road detection in winter road surface conditions using deep neural networks

Improved 3D Object Detector Under Snowfall Weather Condition Based on LiDAR Point Cloud

Recognition of Field Roads Based on Improved U-Net++ Network

Automatic extraction and 3D modeling of real road scenes using UAV imagery and deep learning semantic segmentation

Detection of Road Risk Sources Based on Multi-Scale Lightweight Networks