GPS出行调查中信息自动提取研究进展

onventional household travel survey methods rely on respondents' memory for travel information, which is prone to trip underreporting, time inaccuracy and physical and mental fatigue of respondents. The GPS based survey method is a logical solution to these problems, but its widespread application largely depends on the automation of travel information extraction from the GPS trajectories. This paper reviews and summarizes currently available computer methods for deriving trip ends, trip purpose and travel mode from GPS data. Comparative analyses of these methods indicated that AI-based learning algorithms are best suitable for processing the GPS trajectories with no supplementary data. However, they need to be further improved through referencing empirically-based methods and integrating steps for deriving all three trip elements.