Date of Award
Doctor of Philosophy in Transportation - (Ph.D.)
Civil and Environmental Engineering
I-Jy Steven Chien
Janice Rhoda Daniel
Traffic data collection is one of the essential components of a transportation planning exercise. Granular traffic data such as volume count, vehicle classification, speed measurement, and occupancy, allows managing transportation systems more effectively. For effective traffic operation and management, authorities require deploying many sensors across the network. Moreover, the ascending efforts to achieve smart transportation aspects put immense pressure on planning authorities to deploy more sensors to cover an extensive network. This research focuses on the development and evaluation of inexpensive data collection methodology by using two-dimensional (2-D) Light Detection and Ranging (LiDAR) technology. LiDAR is adopted since it is economical and easily accessible technology. Moreover, its 360-degree visibility and accurate distance information make it more reliable.
To collect traffic count data, the proposed method integrates a Continuous Wavelet Transform (CWT), and Support Vector Machine (SVM) into a single framework. Proof-of-Concept (POC) test is conducted in three different places in Newark, New Jersey to examine the performance of the proposed method. The POC test results demonstrate that the proposed method achieves acceptable performances, resulting in 83% ~ 94% accuracy. It is discovered that the proposed method's accuracy is affected by the color of the exterior surface of a vehicle since some colored surfaces do not produce enough reflective rays. It is noticed that the blue and black colors are less reflective, while white-colored surfaces produce high reflective rays.
A methodology is proposed that comprises K-means clustering, inverse sensor model, and Kalman filter to obtain trajectories of the vehicles at the intersections. The primary purpose of vehicle detection and tracking is to obtain the turning movement counts at an intersection. A K-means clustering is an unsupervised machine learning technique that clusters the data into different groups by analyzing the smallest mean of a data point from the centroid. The ultimate objective of applying K-mean clustering is to identify the difference between pedestrians and vehicles. An inverse sensor model is a state model of occupancy grid mapping that localizes the detected vehicles on the grid map. A constant velocity model based Kalman filter is defined to track the trajectory of the vehicles. The data are collected from two intersections located in Newark, New Jersey, to study the accuracy of the proposed method. The results show that the proposed method has an average accuracy of 83.75%. Furthermore, the obtained R-squared value for localization of the vehicles on the grid map is ranging between 0.87 to 0.89.
Furthermore, a primary cost comparison is made to study the cost efficiency of the developed methodology. The cost comparison shows that the proposed methodology based on 2-D LiDAR technology can achieve acceptable accuracy at a low price and be considered a smart city concept to conduct extensive scale data collection.
Jagirdar, Ravi, "Development and evaluation of low cost 2-d lidar based traffic data collection methods" (2020). Dissertations. 1477.