8/20/2009 5:58:59 PM
 ihagyard Posts: 5
|
I was asked recently to create a contour showing walk distance to a local hospital. On producing my first contour I could easily see that in certain areas the contour was not accurate. This was obvious because I could measure, using MapInfo, from the hospital to points on the contour and comparing the measured distance with the distance used to define the contour. I was using a 25m grid of origin points which I felt would provide a high level of accuracy. The problem was with my connection distance. I was using the standard distance and this was causing problems due to the way in which Accession connects origins to the highway network. Instead of taking the nearest road node to the origin it takes the nearest node in each segment of the circle surrounding the origin. Up to eight nodes in total dependent on connection distance. Then it was building trees from each of those eight points and selecting the shortest path. I found instances where an origin was positioned within 10 metres of a road node yet was connecting to a node 500m away because this provided the shortest walk distance for the overall trip.In reality pedestrians could not walk directly from the origin to that road node. The only way to stop this happening was to reduce the max connection distance. However now my origin points that were further from the road network than this distance became inaccessible and my contour became like swiss cheese and not very coherent. Eventually I found a solution after much trial and error. I exported the road network nodes and imported them as origin points. So now my origin points were each positioned almost exactly on top of a road node. I now set my max connection distance to 10m (the minimum allowed in Accession) and hence had to move my destination points to within 10m of the nearest road node. I selected relaxed interpolation and this provided me with a sensible contour that had very few tiny holes and was accurate all along the road network. Due to the triangulation method utilised by the contouring process in Accession I felt confident looking at the results between my origin points. I wonder whether other people have come across this same issue and whether they found another workaround. Also, I wonder whether Citilabs think it sensible to allow users to choose an option to always select the nearest node to the origin/destination instead of selecting one in each of the eight sectors and then finding the most accessible result.
|
|
permalink
|
8/26/2009 9:24:19 AM
 MJ Posts: 14
|
You may not have a license for it but the Network Analyst tool extension in ArcGIS is infinitely better in producing pedestrian accessibility figures compared to Accession.
Have you tried the 'Increased Road Accuracy' setting in your calculation. If you tick this on you can specify at what interval in metres you want breaks in a link between two nodes, these new breaks you create for the calculation will act as nodes meaning trips from origin points can start their journey from one of these new nodes rather than only snapping to the start/end of link node.
|
|
permalink
|
8/26/2009 9:55:53 AM
 ihagyard Posts: 5
|
Hi MJ. We don't have access to ArcGIS, we use MapInfo exclusively. Good to be aware of though, thanks for that info. I will check out what is involved to obtain a licence.
I did try increasing the road accuracy but this only helps in certain cases, where there are long links. In cases where the link whose nodes you would expect Accession to connect an origin to lies to the south of the origin adding extra nodes doesn't prevent Accession from connecting the origin to a link node that lies to the north-west, say, even if that node is not, in reality, accessible from that origin.
Let me know if what I am describing doesn't make sense and I will try and add a picture.
|
|
permalink
|
8/27/2009 10:12:33 AM
 admin Administrator Posts: 146
|
Ian,
How big is the area you are studying? A more accurate way to get the results out is to use Address Point data, though this will only be useful if doing a smallish calculation, if you use Address point data with Increase Road Accuracy of 100m and a max connection distance of 200m this should give better looking results as all the origin points will be located near the road network at the points of properties on the network,
Regards,
Dan
|
|
permalink
|
9/2/2009 5:30:45 PM
 ihagyard Posts: 5
|
Dan,
The area I was looking at was 3km radius. I will try what you suggest as I can see how it will yield better results than using a grid but I can see that possibly 200m connection distance will still yield inaccurate results in some places where an estate road can connect to a main road less than 200m away by walking in a straight line through an inaccessible area. It may be possible to reduce the connection distance further without making any residences inaccessible to the network. I will test this out when I get chance which may not be in the near future as the project is now complete.
Thanks for your input.
Ian
|
|
permalink
|
Powered by AspNetForum v.6.0.2.0
© 2006-2009 Jitbit Software