Hello Paul, My aim is to make the photo/video mapping process faster and reduce mapper's effort in going through all the photos/videos for information. It becomes very tedious job for mapper to go through several hundreds of photos or hours of long videos to detect signs and integrate to map manually. Plugin will import all the photos/video and search for speed limits and some road signs. A list of photos/video-frame in which features are detected will be displayed to the mapper as well as its position will be displayed on map. When mapper choosed perticular photo/video-frame, he/she will see the detected feature highlighted on photo, it's position on map(according to GPS track provided) and possible key information window prefilled with detected information. Now mapper can verify and edit if required the information/position of feature detected and once everything is ok then mapper can add this feature node and information tag to the map on single click. There will also be many things integrated in the process like displying warning if a feature similar to detected feature already exists in the map etc.
There are several cases which can produce false conclusion, one of such is mentioned by you. Plugin will suggest what it has deteced in image and correspondign possibilities. Solution for such situation as you mentioned can be integrated to get get correct information but it becomes too specific will take good amount of time. There are several factors to be considered like mounting position of camera, distance between camera and object, traffic rules of country you are driving in like right or left driving position etc. Solution for situations like you mentioed can be added later according to specific camera mounting positions, feature which is detected, specific to country etc. My aim is to create a base for photo/video based detection and its integration to map. So I will detect and integrate some basic worldwide signs as a part of GSoC and will try to develope furthur to add specific things as per mapper requirement in GSoC if time permitts or certainly after GSoC. There are almost endless situations like one you have mentioned, one will be able identify such situation and develop a solution to make plugin robust once the base is properly build. Any other suggestions regarding the idea are welcome! -- Nikhil On Wed, Apr 11, 2012 at 2:15 AM, Paul Norman <penor...@mac.com> wrote: > The road signs, max speeds and bus stops don't take long to tag. Take the > bus stop in IMG_3429.JPG as an example. > http://pnorman.dev.openstreetmap.org/photomapping/example_bus_stop.png is > what the area looks like in JOSM, with my car location indicated > > When quickly going through the image I can easily identify that there is a > bus stop here. Seeing that there is a bus stop, I first have to identify > where it is. Neglecting GPS error, the bus stop is about 50m away from my > car and on the other side of the road. As this bus stop has a shelter I can > easily see it on my imagery. If it didn't, I would have to count power > poles, look at curves in the road or some other technique to get the precise > location. Even if I had no imagery I would want to get the stop on the > correct side of the road. > > Suppose there is no bus_stop in OSM. I then need to add a node to where it > is with highway=bus_stop. I then look at the bus stop and see that it has a > shelter, bench, is wheelchair accessible, has a waste basket and is operated > by TransLink. I then convert these to shelter=yes bench=yes wheelchair=yes > amenity=waste_basket operator=TransLink. If the image was high enough > quality I would add ref=* but my camera can't make it out from 50m away. > > To simply create a highway=bus_stop at my position would be wrong. I'm > driving on the right side of the road and the bus stop is on the left. Even > identifying the bus_stop doesn't help much, that's the easiest part of > adding it. > > How would your proposed plugin help me? > >> -----Original Message----- >> From: Nikhil Upadhye [mailto:nikhil.spitf...@gmail.com] >> Sent: Tuesday, April 10, 2012 6:52 AM >> To: talk@openstreetmap.org >> Subject: Re: [OSM-talk] An example photo/video mapping trip >> >> Thanks Paul for sharing files! >> I have submitted proposal for JSOM plugin development to extract >> information from such images and videos under GSoC this year. >> This plugin will use image processing to extract the key features such >> as road signs, max speed, bus stop etc as you have indicated in some >> images. These extracted features from images/videos will be >> automatically added to map with respective tags after synchronizing with >> gps tracks. >> >> I would request if anyone has such images/videos along with gpx file to >> share here as these will be helpful while developing the plugin. >> >> -- >> Nikhil Upadhye >> >> >> >> On Tue, Apr 10, 2012 at 7:11 AM, Paul Norman <penor...@mac.com> wrote: >> > A few GSoC students have made proposals relating to video and photo >> mapping. >> > As I make good use of photos from my car for mapping I thought I'd >> > prepare a set of annotate photos, pointing out objects that I map. >> > >> > I've finished these and placed them at >> > http://pnorman.dev.openstreetmap.org/photomapping/example_photomap_hig >> > hway.z >> > ip >> > The zip file includes the images, the GPX track and some notes. The >> > image files have coloured rectangles indicating different objects with >> > the key in the notes. >> > >> > The trip is a mix of residential, industrial and highway travel. Most >> > of the time spent entering the data from this trip would normally be >> > typing out the tagging for POIs and shops. >> > >> > My typical trip involves much less highway travel and much more travel >> > through retail and commercial landuse where I can spend a couple >> > minutes with one picture adding all the shops. >> > >> > The photos were taken 10 seconds apart from a camera mounted in about >> > the middle-right of my car and my GPX track comes from my garmin. >> > >> > >> > _______________________________________________ >> > talk mailing list >> > talk@openstreetmap.org >> > http://lists.openstreetmap.org/listinfo/talk >> >> _______________________________________________ >> talk mailing list >> talk@openstreetmap.org >> http://lists.openstreetmap.org/listinfo/talk > _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk