Map matching tracking data

Abstract

Route planners become more and more available for everyone. These intelligent devices are very popular and are very easy in use. The police of Ghent had a problem with new and transferred cops. They did not know the environment, so it would be better if they got a route planner in their car which showed the right way. The problem is that the ordinary route planners are not able to fulfill all the wishes for law enforcement. They do not take into account for instance where the school gate is situated and to avoid this area when they are on an intervention around school time if possible. This thesis tries to make a custom-made route planner for police officers. This thesis is about mapping the GPS points got from real-life trajectories of experienced police officers mapped to the real network topology. The analysis of these trajectories afterwards is very interesting and will be done in further work. How do they get to the place of intervention? Do they take many risks by driving too fast or are they careful enough and do not drive too fast at all. We are trying to make a custom-made route planner for these officers especially for new or transferred cops who do not know the environment yet. We did this by analysing real trajectories from police officers on patrol. A first step is to record these real-life trajectories. Then we have these trajectories we can map to the real network topology, analyze it and finally we can make a custom-made route planner. In this work we try to map the GPS points recorded to the real network topology. We try a relatively new technique called beads. This is an estimate where the police car could have been between two succeeding points knowing that they do not exceed some maximum speed and knowing the time between both points. Our recordings are pretty accurate in theory (every 10 meters a GPS point) but in practice this is of course not always the case. Sometimes there are even gaps of more than 400 meters. Most algorithms found in literature are based on real-time mapping and they have points time-based recorded for instance every 5 seconds a point. These are the biggest differences with our data. The fact that we have every 10 meters a point recorded is in our advantage because then our beads algorithm performs best. It can then better calculate where it could have been between two succeeding points, than if it was time-based recorded.