Date of Award
Doctor of Philosophy
Civil, Construction, and Environmental Engineering
Civil Engineering(Intelligent Infrastructure Engineering)
Accurate traffic data collection is essential for supporting advanced traffic management system operations. This study investigated a large-scale data-driven sequential traffic sensor health monitoring (TSHM) module that can be used to monitor sensor health conditions over large traffic networks. Our proposed module consists of three sequential steps for detecting different types of abnormal sensor issues. The first step detects sensors with abnormally high missing data rates, while the second step uses clustering anomaly detection to detect sensors reporting abnormal records. The final step introduces a novel Bayesian changepoint modeling technique to detect sensors reporting abnormal traffic data fluctuations by assuming a constant vehicle length distribution based on average effective vehicle length (AEVL). Our proposed method is then compared with two benchmark algorithms to show its efficacy. Results obtained by applying our method to the statewide traffic sensor data of Iowa show it can successfully detect different classes of sensor issues. This demonstrates that sequential TSHM modules can help transportation agencies determine traffic sensors’ exact problems, thereby enabling them to take the required corrective steps.
The second research objective will focus on the traffic data imputation after we discard the anomaly/missing data collected from failure traffic sensors. Sufficient high-quality traffic data are a crucial component of various Intelligent Transportation System (ITS) applications and research related to congestion prediction, speed prediction, incident detection, and other traffic operation tasks. Nonetheless, missing traffic data are a common issue in sensor data which is inevitable due to several reasons, such as malfunctioning, poor maintenance or calibration, and intermittent communications. Such missing data issues often make data analysis and decision-making complicated and challenging. In this study, we have developed a generative adversarial network (GAN) based traffic sensor data imputation framework (TSDIGAN) to efficiently reconstruct the missing data by generating realistic synthetic data. In recent years, GANs have shown impressive success in image data generation. However, generating traffic data by taking advantage of GAN based modeling is a challenging task, since traffic data have strong time dependency. To address this problem, we propose a novel time-dependent encoding method called the Gramian Angular Summation Field (GASF) that converts the problem of traffic time-series data generation into that of image generation. We have evaluated and tested our proposed model using the benchmark dataset provided by Caltrans Performance Management Systems (PeMS). This study shows that the proposed model can significantly improve the traffic data imputation accuracy in terms of Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) compared to state-of-the-art models on the benchmark dataset. Further, the model achieves reasonably high accuracy in imputation tasks even under a very high missing data rate (>50%), which shows the robustness and efficiency of the proposed model.
Besides the loop and radar sensors, traffic cameras have shown great ability to provide insightful traffic information using the image and video processing techniques. Therefore, the third and final part of this work aimed to introduce an end to end real-time cloud-enabled traffic video analysis (IVA) framework to support the development of the future smart city. As Artificial intelligence (AI) growing rapidly, Computer vision (CV) techniques are expected to significantly improve the development of intelligent transportation systems (ITS), which are anticipated to be a key component of future Smart City (SC) frameworks. Powered by computer vision techniques, the converting of existing traffic cameras into connected ``smart sensors" called intelligent video analysis (IVA) systems has shown the great capability of producing insightful data to support ITS applications. However, developing such IVA systems for large-scale, real-time application deserves further study, as the current research efforts are focused more on model effectiveness instead of model efficiency. Therefore, we have introduced a real-time, large-scale, cloud-enabled traffic video analysis framework using NVIDIA DeepStream, which is a streaming analysis toolkit for AI-based video and image analysis. In this study, we have evaluated the technical and economic feasibility of our proposed framework to help traffic agency to build IVA systems more efficiently. Our study shows that the daily operating cost for our proposed framework on Google Cloud Platform (GCP) is less than $0.14 per camera, and that, compared with manual inspections, our framework achieves an average vehicle-counting accuracy of 83.7% on sunny days.
Huang, Tongge, "Designing the next generation intelligent transportation sensor system using big data driven machine learning techniques" (2020). Graduate Theses and Dissertations. 18143.