Loading Situation Visual Analysis Based On Taxi Trajectory Data

Urbanization promotes the development of social economy and improves people’s living standard, meanwhile it brings serious problems such as traffic jams. In order to solve these problems, more sensors like GPS devices are mounted on a taxi. These GPS devices allow people to collect massive taxi trajectory data.Taxi trajectory data contains a great quantity of knowledge which we can use to analyze people’s travel behaviors, optimize the city traffic and improve the travel condition. However, trajectory data itself is relatively large, complex, and difficult for us to understand. The technology of visual analysis provides an effective way to show and analyze the data. In this thesis we use taxi trajectory data to analyze the people’s travel hot spots. Then with these hot spots, we try to help taxi drivers to find the optimal driving track to pick up passenger. At last we develop a visual prototype system to support intuitive visual analysis of hot spots and driving tracks. The main contents of this thesis include the following aspects:(1) Taxi trajectory data preprocessing. By setting trajectory extracted rules and methods,we filtrate jump points, the parking trajectory and the too short trajectory from taxi trajectory data to obtain a real taxi trajectory.(2) The hot spots extraction and analysis. An improved clustering algorithm-GBADBSCAN is proposed to generate hot spots. After getting passengers up and down points based on the pre-extraction of taxi trajectory data, we use GBADBSCAN to generate hot spots. Then a visualization method is designed to analyze the distribution of hot spots using the passengers up and down cluster icon.(3) Get a recommended taxi running trajectory. Firstly,this thesis adopts a similar track classifed method based on the beginning and end points to divided all trajectories into different track subsets so that the trajectories in same subset have the hear beginning and end points. Secondly, we use a density-based ε distance trajectory clustering algorithm to cluster trajectories for identifying candidates. At last we built a weighted tree of the candidate trajectories to find the optimal recommended trajectory for the taxi driver to pick up passenger in a nearby hot spot.(4) Implementing a visualization prototype system. Five components of map, time, parallel coordinates, console, and navigation path description have been adopted to built a visualization prototype for the visualization of the taxi trajectory data.