Data-driven parallelizable traffic incident detection using spatio-temporally denoised robust thresholds

Abstract

Automatic incident detection (AID) is crucial for reducing non-recurrent congestion caused by traffic incidents. In this paper, we propose a data-driven AID framework that can leverage large-scale historical traffic data along with the inherent topology of the traffic networks to obtain robust traffic patterns. Such traffic patterns can be compared with the real-time traffic data to detect traffic incidents in the road network. Our AID framework consists of two basic steps for traffic pattern estimation. First, we estimate a robust univariate speed threshold using historical traffic information from individual sensors. This step can be parallelized using MapReduce framework thereby making it feasible to implement the framework over large networks. Our study shows that such robust thresholds can improve incident detection performance significantly compared to traditional threshold determination. Second, we leverage the knowledge of the topology of the road network to construct threshold heatmaps and perform image denoising to obtain spatio-temporally denoised thresholds. We used two image denoising techniques, bilateral filtering and total variation for this purpose. Our study shows that overall AID performance can be improved significantly using bilateral filter denoising compared to the noisy thresholds or thresholds obtained using total variation denoising. AbstractAutomatic incident detection (AID) is crucial for reducing non-recurrent congestion caused by traffic incidents. In this paper we propose a data-driven AID framework that can leverage large-scale historical traffic data along with the inherent topology of the traffic networks to obtain robust traffic patterns. Such traffic patterns can be compared with the real-time traffic data to detect traffic incidents in the road network. Our AID framework consists of two basic steps for traffic pattern estimation. First, we estimate robust univariate speed threshold using historical traffic information from individual sensors. This step can be parallelized using MapReduce framework thereby making it feasible to implement the framework over large networks. Our study shows that such robust thresholds can improve incident detection performance significantly compared to traditional threshold determination. Second, we leverage the knowledge of the topology of the road network to construct threshold heatmaps and perform image denoising to obtain spatio-temporally denoised thresholds. We used two image denoising techniques, bilateral filtering and total variation for this purpose. Our study shows that overall AID performance can be improved significantly using bilateral filter denoising compared to the noisy thresholds or thresholds obtained using total variation denoising.

A Hybrid Method for Traffic Incident Detection Using Random Forest-Recursive Feature Elimination and Long Short-Term Memory Network With Bayesian Optimization Algorithm

“…In some studies, DR, FAR and MTTD are used to construct Performance Index (PI) for performance evaluation. In this study, the PI proposed in Reference [15] is used to evaluate the performance of AID methods. The calculation formula of this PI is as follows.…”

“…Threshold determination still attracts attention. Recently, an AID method using spatiotemporally denoised robust thresholds has been proposed, and results show that these robust thresholds can improve incident detection performance significantly compared to traditional threshold determination [15] Since the 1990s, there have been some AID methods based on machine learning, such as Artificial Neural Network (ANN) [16], [17], [18], Support Vector Machine (SVM) [19], [20], [21], [22], [23], decision trees [24] and Bayesian classifiers [25]. For these AID methods, traffic incident detection is regarded as a pattern classification problem, more precisely, a binary-classication problem (incident or non-incident).…”

Detecting and interpreting non‐recurrent congestion from traffic and social media data

“…signal control, temporary street parking), the average speed of each section is calculated in a 15‐min non‐overlapping interval, similar to the work of Chakraborty et al. [7]. Thus, the length of the time‐of‐day equals .…”

Anomaly detection and classification in traffic flow data from fluctuations in the flow-density relationship

“…This approach is taken from Chakraborty et al (2019), where one avoids 'swamping' of the robust statistics. As in the cited work, we set the threshold values s max 1 , s max (2005).…”