Generating a Vehicle Trajectory Database from Time-Lapse Aerial Photography

Abstract

The advancing art of microsimulation modeling is driving a growing interest in vehicle trajectory databases. These can be rich sources of travel information, providing origin–destination patterns, path choices, travel times or delays, and lane changing behavior. These metrics have historically been difficult to acquire, and even with the approach of the big-data era, it remains a challenge to collect high-quality, granular vehicle trajectory data. This paper describes how use of state-of-the-practice time-lapse aerial photography (TLAP), acquired continuously for up to several hours at 1-s frame rates, can produce trajectory data sets with high granularity. These surveys involve the use of airborne digital cameras held in stationary positions about one mile above the ground, to record the movement of virtually all highway vehicles in defined study areas. In this paper, TLAP’s differences from similar military imaging capabilities, such as wide-area motion imagery, are discussed, to explain how TLAP can be affordable for highway traffic studies. Then local survey design and execution for a study in the Phoenix, Arizona, area are presented. Flight and photography survey planning, image alignment, and data extraction methods are discussed; and study output and graphics are presented near the end. Big data someday will likely be magically able to supply all the field data needed for complex analysis or simulation of specific study areas; for now, however, 1-s TLAP can provide many of those benefits.

On the new era of urban traffic monitoring with massive drone data: The pNEUMA large-scale field experiment

“…Additionally, aerial observations are examined to study moving synchronized flow patterns by measuring trajectories for all vehicles. In (Zhang et al, 2016), although no UAVs were used, researchers used manned helicopters and Time-Lapsed Aerial Photography (TLAP) to create a large dataset for monitoring drivers behavior. Using two helicopters equipped with seven cameras in total covered an almost 5 square miles grid area.…”