User:Matt

From OpenStreetMap

I've been almost contributing to this project since it began, but I'm too good at my other hobby (procrastination) to have actually Gotten Things Done, as Steve would say!

Anyway, it should be possible to procrastinate more efficiently on the Wiki ;-)

Currently working on...

GPS tracking is too hard in built-up areas: I don't often get a lock and when I do there are massive multi-bounce errors, making the data pretty worthless. The errors are compounded by the streets in urban areas being closer together and often narrower. There are also more points of interest such as post-boxes and bus stops. The aerial imagery is pretty good (in London), but often roads are obscured or difficult to trace accurately due to perspective artefacts or tree cover. I could carry around lots of "real" surveying equipment, but I'm lazy and would prefer a single instrument to do all the work.

SLAM (Simultaneous Location and Mapping) is a technique used in robotics to use feature-identification on video frames to provide a good idea of where the robot is. This paper is a good place to start and there is some software, but I wasn't able to get it to work. The plan would be to walk around until the camera's memory card is full and then post-process that into a 2D plan, georectify it based on known points (e.g: selected from aerial imagery or existing mapped roads) and sketch ways on top of that. A big advantage would be the ability to inspect every frame for road signs, street names, bus stop numbers, etc...

Current status: pre-alpha.