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.

Application of Neural Network in the Data Fusion

Multimedia data fusion method based on wireless sensor network in intelligent transportation system

Neural Network-Based Freeway Traffic State Data Fusion System Using Cellular Handoff Probe System and Microwave Sensors

Urban Traffic Controller Using Fuzzy Neural Network and Multisensors Data Fusion

Data Fusion Method for Traffic Information Detected by Fixed Detectors and Floating Cars

An approach for real-time urban traffic state estimation by fusing multisource traffic data

Urban Fine Management of Multisource Spatial Data Fusion Based on Smart City Construction

Traffic Data Fusion Research Based on Numerical Optimization BP Neural Network

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

Freeway Multisensor Data Fusion Approach Integrating Data from Cellphone Probes and Fixed Sensors.

A Fuzzy Fusion Algorithm to Identify the Traffic States by the Floating Car Data.

An Urban Traffic Speed Fusion Method Based On Principle Component Analysis And Neural Network

Study on data fusion model with multi-source heterogeneous traffic data

Optimization Model of Urban Rail Transit Operation Management Based on Neural Network Algorithm

Estimation of urban arterial real-time travel speed based on multi-source data fusion

An Approach to Urban Traffic State Estimation by Fusing Multisource Information

Traffic Flow Prediction Base on Feature Fusion of Multi-modal Data

An Applications of Neural Network Information Fusion Techniques in Fault Detection and Diagnosis

Traffic Incident Detection for Urban Arterial Road Based on Data Fusion and Learning Vector Quantization

Traffic Control Based on Data Fusion

Data Fusion for Multi-Source Sensors Using GA-PSO-BP Neural Network