Innovative Procedures for Travel Data Collection and Processing

Abstract

Global Positioning System (GPS) or Smartphone technology has been increasingly used in travel data collection. Although GPS devices can directly record spatial and temporal information, trip ends, travel modes and trip purposes are not recorded. So GPS data processing becomes a critical procedure to produce these results, which can be used in transport planning. It has been proved that GPS records are more reliable than travel diaries; however, the quality of GPS data processing work usually influences the quality of results. Researchers have been engaging in the improvement of GPS data processing for the past decade. Traditionally, data processing for GPS records (from dedicated GPS loggers and Smartphones) includes three steps, namely trip identification, mode detection and purpose imputation. However, the results of mode and purpose detection are entirely based on the result of trip identification. Hence, the total accuracy of a GPS survey would be the product of the accuracy of each step. This thesis focuses on the improvement of travel data quality by improving data collection and processing. In this study, a new procedure is introduced which combines the process of trip identification and mode detection. Some general rules (i.e., a threshold of dwell time and the time interval for recording data) are tested. This research also firstly applies a new technology, a life-logging camera, to travel data collection. Images are used to help to pursue ground truth -- especially recorded trips in which GPS data were missing -- and detect some types of travel modes in order to improve the accuracy of data processing. An automating image processing procedure is proposed and tested in this study. In addition, a concept of “mode-point-chain” is discussed to identify the cases of mode change and modify incorrect mode detection results. For the process of purpose imputation, more travel information is suggested to be used in the process. This thesis also uses tour-based information in trip purpose imputation to improve the results. By using the new procedure, the trip identification accuracy was increased by almost 30 percent, taking the missing trips into account. Since trip identification and mode detection were combined, this increase also benefits mode detection results. With the help of image processing and the new procedure of mode change detection, the accuracy of mode detection increased by 7% regardless of the accuracy increase in trip identification. The new processing method also increased the accuracy of trip purpose imputation by 8%. This improvement can help researchers and planners obtain more accurate data for decision making and planning.