Subscribe to this thread
Home - General / All posts - Normalize topology with drawing
Ian
215 post(s)
#07-Jan-19 01:22

I have a paddock drawing that has been supplied to me - just a drawing with polygons - whoever did it was pretty sloppy and a lot of the nodes haven't been snapped together. I need to edit the drawing so it sits correctly over some new imagery. I started using shared edit but obviously with the nodes not snapped together it is often not working. To sort the issue out I have done a normalize topology with a precision of 0.5 then 1m. The problem I am having is that when I perform the transform it isn't getting all of the points. It corrects some but not all so I keep coming across points that haven't been joined together. The precision isn't the issue as when I re-do the transform using the same setting it snaps them together. My questions are; Why wouldn't the transform get all of the points at once (are there other issues that can affect this operation)? And if it is standard to have to redo it how do I know when I have got them all?

My process is view-properties-precision - 1.0m. In the transforms tool bar, all objects (though have also tried using selection) - normalize topology - apply. Have done this in the drawing window and also in the map window with the paddock layer highlighted.

tjhb

8,474 post(s)
online
#07-Jan-19 05:06

The iteratively better results are explicable given that each pass starts with adjusted data. New data, new questions, new answers.

How to solve definitely in one pass? Can’t.

How to know whether everything is solved? Basically, can’t. (But custom SQL could help here.)

How to ensure that each successive pass only makes improvements, no regressions? Absolutely can’t. Any given pass might destroy the detail in your data, what you really care about.

There really is no substitute for people doing things correctly in the first place. Cheap data is really expensive.

tjhb

8,474 post(s)
online
#07-Jan-19 08:55

Given that a quick and reasonable attempt to correct the data has not completely worked, I would reject it and ask for it to be re-done competently, if you have that option.

Ian
215 post(s)
#08-Jan-19 00:44

Not the answer I was hoping for hoping I had just missed something and there was a quick fix. Can't get it re-done have to suck it up and do it manually unfortunately.

tjhb

8,474 post(s)
online
#08-Jan-19 06:07

A quick fix might be possible but would require time and cost (and data).

tjhb

8,474 post(s)
online
#08-Jan-19 06:12

Another free thing that might be worth trying: Orthogonalize.

Do that at double the target location precision, then normalize topology once.

No guarantees, but try that.

tjhb

8,474 post(s)
online
#08-Jan-19 06:44

So if your target location precision is 1m, try first orthogonalizing at 0.5m (or perhaps less, e.g. 0.1m).

tjhb

8,474 post(s)
online
#08-Jan-19 07:24

P.s. why can't you ask for the data to be redone?

If it was not done competently, why not ask for a resupply?

dchall8
530 post(s)
#08-Jan-19 14:56

why can't you ask for the data to be redone?

In my case, my predecessor died. The interim mappers were either woefully in over their heads or they were ESRI contractors who were either unwilling or unable to get the accuracy and precision necessary for the job. Here is an example that I have been dealing with since 2014.

This neighborhood has 1,000 blocks with 24 lots each. They are all identical and aligned with the compass. What could be easier? For some reason, nobody thought to create a master drawing of the subdivision and use that as a guide. Instead they hand drew each parcel, apparently guessing at the distances. The image shows a corrected block (top) and the uncorrected (lower) with the master showing through in blue. These make me seasick before I correct them. So have been correcting these as I come across them for years with no end in sight.

Attachments:
Avalon errors.jpg

artlembo

3,101 post(s)
#08-Jan-19 17:39

what about using the rubbersheeting capability in Manifold 8 to get them even closer?

tjhb

8,474 post(s)
online
#08-Jan-19 18:11

...Or doing it in one hit in SQL, an easy case if it is all so close to the master drawing.

(Not so easy in Ian's case. But possible in principle.)

geozap118 post(s)
online
#08-Jan-19 18:21

If lots are identical and aligned, woulnd't that be relatively easy to recreate them from scratch and then transfer any useful existing data from the problematic ones to them?

"For some reason, nobody thought to create a master drawing of the subdivision and use that as a guide." Maybe they did automatic raster-to-vector conversion.

dchall8
530 post(s)
#09-Jan-19 20:39

geozap, now you have me thinking. There are 3,700 parcels in the area. Are you suggesting there is a way to go from the legal description of the parcels (with lots and block numbers) to being able to select just the correctly numbered parcels from the master and creating a new layer with the parcels formed by a union into the correct shape and location??? An example of one of the more complicated parcels is AVALON C BLK 96 LT 6-10 21-24 0.465 ACRES. Removing the last two tokens leaves the pertinent information to form the parcel. Each parcel contains fields for the subdivision (Avalon C), block, and lot. I would think the dash notation, 6-10, would present a problem for automating directly, but I could probably come up with a way to expand that. Not sure of the syntax needed to make this work, though.

Art, not sure if you were replying to me, but I had not thought of convex hulls to redo these parcels. I tried two approaches to it. First I tried selecting the appropriate parcels from the master and using the Convex hull transform to create a new parcel in the correct layer. Starting with zero muscle/mousle memory I made a lot of mistakes the first few times. I found myself forgetting to select the new layer after selecting the parcels in the master layer. So I went with that approach and then copied all the new parcels into the working layer afterwards. What I did not care for was the creation of a point at every corner of every lot. The second approach was to create points snapped to the corners using the master lot layer to snap to. New points went into another layer. Then I collected the appropriate points and ran the Convex hull transform on the points. Then I moved the new parcels into the working parcel layer to renumber them. This alleviated the problem of extraneous points at the corners of the lots. Between the two approaches I like the second better. Between this and my old approach of moving/snapping all the points to the correct location I think I like Convex Hulls better. There was a lot of having to aim with the mouse in the old approach, and, as my eyes are aging, the Convex Hull transform seems to be moderately less hassle. My predecessors sometimes created multiple points at the corners which added a hassle factor to the old way. Now I can simply delete the old parcels and not deal with their quirks.

I wonder if Convex Hulls would help Ian with his issue? It's not automatic, but it might be better than manually moving all the points.

artlembo

3,101 post(s)
#09-Jan-19 20:59

Not convex huls, but rather rubber sheeting. Just thinking of a way to re-adjust the data.

If you wanted to send me some of your data that I could have a look at it. I’m in a little bit of a downtime right now before the semester starts back up

tjhb

8,474 post(s)
online
#09-Jan-19 21:33

A.k.a georegistration, right Art? I had not heard it called rubber sheeting before either.

artlembo

3,101 post(s)
#09-Jan-19 22:26

Yes, oops. I’m reverting back to my GIS vocabulary :-)

Georegistration, but instead of Conformal, Affine, or Polynomial, use Triangulation which is rubbersheeting.

geozap118 post(s)
online
#10-Jan-19 10:06

I had in mind recreating the parcels using some cad software using cad functions to copy-pasting parcels, or offsetting their sides, etc., maybe also using some autolisp to auto-create numbering of parcels, and then importing to manifold to create areas.

I am not familiar with US legal descriptions, but if legal descriptions provide information about parcels size and position systematically and with strict integrity, using that information (as you write, if get that right) seems even better.

Manifold User Community Use Agreement Copyright (C) 2007-2017 Manifold Software Limited. All rights reserved.