Feasibility of Estimating Commodity Flows on Highways with Existing and Emerging Technologies
Document Type
Report
Publication Date
2019
Pages
31 pp.
Abstract
Understanding the types and volumes of different commodities being hauled over the roadway network is important for freight demand modeling and planning. Categorizing trucks or trailers into subclasses (e.g., container, dry van, refrigerated vans, tank, and car transporter) will help narrow down the type of commodity that is being carried. For example, refrigerated trailers are commonly used to transport perishable produce and meat products, tank trailers are for fuel and other liquid products, and livestock is carried in specialized trailers. The main goal of this project is to investigate the feasibility of using data from non-intrusive sensors (e.g., camera and LIDAR) to identify the type of trailer. Algorithms are developed to demonstrate the feasibility of accurately detecting and classifying the trailer types. Data collected by a 3D LIDAR sensor are analyzed to extract pertinent information for classifying truck trailers into subtypes. The collected data include enough samples of four trailer types: Intermodal container, refrigerated container, dry van, and refrigerated dry van. After data processing, various machine learning algorithms are developed to automatically classify the observed trucks into these four subcategories. The tested machine learning algorithms include an SVM (support vector machines) model which requires a feature vector as the input, and deep CNN (convolutional neural networks) which do not require extracting features from raw data. Instead, LIDAR data are converted to a 2D image as the input to the CNN. The results from these machine learning algorithms show that these four truck body types could be classified with over 90% accuracy. The report present further details on how raw data are processed and the classification algorithms are developed.
Rights
"No restrictions. This document is available from the National Technical Information Service, Springfield, VA 22161."
ORCID
0000-0003-2003-9343 (Cetin), 0000-0001-5251-0748 (Vatani)
Original Publication Citation
Cetin, M., Sahin, O., Vatani, R. N., Zatar, W., & Nichols, A. P. (2019). Feasibility of estimating commodity flows on highways with existing and emerging technologies. United States Department of Transportation. https://rosap.ntl.bts.gov/view/dot/44338
Repository Citation
Cetin, Mecit; Sahn, Olcay; Vatani, Reza Nezafat; Zatar, Wael; and Nichols, Andrew P., "Feasibility of Estimating Commodity Flows on Highways with Existing and Emerging Technologies" (2019). Civil & Environmental Engineering Faculty Publications. 84.
https://digitalcommons.odu.edu/cee_fac_pubs/84