Multi-Stage Approach to Travel-Mode Segmentation and Classification of GPS Traces

Abstract

ABSTRACT:This paper presents a multi-stage approach toward the robust classification of travel-modes from GPS traces. Due to the fact that GPS traces are often composed of more than one travel-mode, they are segmented to find sub-traces characterized as an individual travel-mode. This is conducted by finding individual movement segments by identifying stops. In the first stage of classification three main travel-mode classes are identified: pedestrian, bicycle, and motorized vehicles; this is achieved based on the identified segments using speed, acceleration and heading related parameters. Then, segments are linked up to form sub-traces of individual travel-mode. After the first stage is achieved, a breakdown classification of the motorized vehicles class is implemented based on sub-traces of individual travel-mode of cars, buses, trams and trains using Support Vector Machines (SVMs) method. This paper presents a qualitative classification of travel-modes, thus introducing new robust and precise capabilities for the problem at hand.

Travel Mode Detection Based on GPS Raw Data Collected by Smartphones: A Systematic Review of the Existing Methodologies

“…The reference lists of excluded reviews were reviewed and potential papers were gathered. Finally, 12 published papers matching all the criteria were included in this review [1,31,[38][39][40][41][42][43][44][45][46][47], as shown in Table 2. In particular, we selected [39] because the segment identification method of this study is a significant data preprocessing technique for travel mode detection.…”

“…Zhang et al [44] calculated the values of speed and heading from point positions and time stamps and utilized a smoothing method to reduce speed errors by averaging the neighborhood.…”

“…Zhang et al [44] derived individual travel-mode segments from GPS traces by identifying stops. As they mentioned, very low speed and very small distance changes could be defined as stops.…”

“…Zhang et al [44] presented a novel multi-stage method to identify travel modes. In the first stage, they used three parameters (mean speed, maximum speed, and heading related changes) to identify the three main travel modes (walking, bicycles, and motorized vehicles) on the basis of the identified segments.…”

Classifying Human Activity Patterns from Smartphone Collected GPS data: A Fuzzy Classification and Aggregation Approach

“…These limitations increase the difficulty in extracting health‐related activity information from smartphone‐measured GPS traces. Current protocols of GPS data processing were primarily designed for transportation mode detection (Gonzalez et al , Zheng et al , Zhang et al , Gong et al ), travel behavior analyses (Wolf et al , Auld et al , Papinski et al , Schüssler and Axhausen , Clark and Doherty , Oliver et al , Millward and Spinney , Millward et al ), and movement analyses (Ashbrook and Starner , Vazquez‐Prokopec et al ). A common practice of these protocols is to divide the sequence of GPS points into segments based on similarities of point characteristics such as speed and acceleration.…”