Maximizing Port and Transportation System Productivity by Exploring Alternative Port Operation Strategies

2011PDFResearch PaperGTI-09-03PortsTruck trafficGatesAlgorithmsService timeTraffic congestionPort operationsCamerasTraffic surveillanceHarborsTruckingGatesTraffic monitoringAlgorithmsTraffic congestionCamerasGeorgiaGeorgia Transportation Institute University Transportation CenterTsai, JamesWu, YichingShen, JunanGeorgia Transportation InstituteGeorgia Institute of TechnologyUnited States. Department of Transportation. Research and Innovative Technology AdministrationNTL-FREIGHT-Trucking IndustryNTL-MARINE/WATERWAYS TRANSPORTATION-PortsNTL-OPERATIONS AND TRAFFIC CONTROLS-CongestionUS Transportation CollectionSeaports are a critical transportation component that supports the nation\u2019s economy. Many U.S. ports are now experiencing significant truck congestion at the gate, which decreases the productivity of ports and truck fleets (e.g. truck wait times) and increases vehicle exhaust emissions, which contributes to air pollution. Actual truck traffic data at the gate, including arrival time, service/processing time, and wait/queue time, is essential for studying truck congestion, but such data has been difficult to obtain with existing manual data collection methods. This research proposes a service time extraction algorithm using video log images taken by surveillance cameras at the gate to effectively acquire this much-needed data. A service time extraction algorithm consisting of three unique components, 1) a design of two lane-based regions of interest (ROIs) to represent truck trajectories, 2) a frame-differencing change detection algorithm addressing low frame-rate and cast shadow issues, and 3) a unique transition model with a set of decision rules that considers perspective occlusion and other potential sources of false positive detections, was developed to reliably detect truck departures. The performance of the proposed algorithm was evaluated using 6,567 actual images captured via internet at a low frame-rate from a live video feed from a gate surveillance camera in the U.S. Preliminary results demonstrate the robustness of the proposed algorithm by successfully detecting truck departures under various challenging conditions, including day and night lighting conditions, perspective occlusion, cast shadows, multi-lane departures, and non-truck movements. The algorithm achieved a correct detection rate of 98.1% for all the images, which can sufficiently represent truck service times at a gate. To further extend the use of this vision-based technology, a vision-based, multi-view gate data acquisition module is proposed to collect the images at the Port of Savannah for wait time extraction validation. In addition, the Georgia Institute of Technology, the Center of Innovation for Logistics, and private sector corporations have initiated a project to extend the proposed algorithm for monitoring and optimizing the flow of truck traffic in the roadway network near the Port of Savannah.92