| Traffic violations are direct causes of traffic accidents,undermining human safety and causing economic losses.The clustering of traffic violations leads to traffic violation hotspots in road networks where traffic enforcement strategies can be implemented to mitigate the traffic violation issue.The development of smart cities brings spatial crowdsensing big data,providing new opportunities to inspect traffic violation hotspots.Therefore,we propose a framework for traffic violation hotspot identification,diagnosing,and prediction based on spatial crowdsensing big data,detailed as follows:First,we identify traffic violation hotspots based on crowd behavior analysis.We first normalize the vehicle trajectories with map-matching algorithms to extract driving behaviors and then model driver perspectives to restore spatiotemporal contexts of driving behaviors and identify traffic violations.Finally,we extract spatiotemporal patterns of traffic violations to get the traffic violation hotspot distribution.Second,we diagnose traffic violation hotspots based on cross-domain data fusion.We first select typical traffic violation hotspot locations by clustering algorithms and then build a driving simulator using road environments and point clouds.Driving simulation tests are conducted to generate indicators for diagnosing traffic violation hotspots.Finally,a causality diagnosing model is built to diagnose traffic scenes automatically.Third,we predict traffic violation hotspots based on spatiotemporal context modeling.We first extract spatiotemporal contexts of traffic violation hotspots from road environments and then combine tri-training and active learning to predict traffic violation hotspots.Finally,we optimize patrol routes based on the prediction results.We evaluate our methods on real-world spatial crowdsensing big data and conduct case studies.The results show that our framework identifies,diagnoses,and predicts traffic violation hotspots effectively and comprehensively.The framework can help urban authorities understand dynamic urban traffic violation hotspots,improve traffic infrastructure,and optimize labor and non-labor resource allocation and scheduling. |
