Site Prep

Title: Agree 100%

By: Barry J. Theodore

Posted: March 11, 2010 3:47 PM

I read the above article by you. I agree with you 100% that there needs to be a new method for 3D earthwork. The TIN surfaces need to be smoothed similar to the way the contours are, to reflect the surfaces in the real world. There are hard surfaces, such as asphalt and concrete, and there are soft surfaces, such as loam and dirt.

Maybe a "surface curvature" application to the soft surfaces could be developed, similar to, or the same as the "soft breaklines" you had mentioned. As I recall from calculus, curvature was equal to the inverse of the circumference of a curve, i.e. 1/c. Perhaps there could be similar factors that would produce the desired smoothing that is required.

Title: Keep it simple

By: Louis Spahn

Posted: March 11, 2010 3:49 PM

3D Grading is a core thing of mine. Keeping the process simple and accurate is the Base that it is built on.

After a quick scan of the article, a lot of what you are asking for is present in Agtek. Some of the things asked for nearer the bottom of the article, seem to be from computer based people.

John Fletcher the owner of Agtek did want me to pass on Agtek’s motto... “Dirt Simple Solutions”... Escarpments, Caves, and Flowerpots, are a very, very miniscule part of the grading volume calculation world. To get the program to understand these kinds of features would add a great deal of overhead to the programs, for very little improvement for most of the users.

Convex Hulls are easy for the computer to deal with but Concave Hulls are a very different animal; that is what would be required to process Escarpments, Caves, and Flowerpots.

Title: No hurry to change

By: Donald Nicholson

Posted: March 11, 2010 3:50 PM

There actually has been software with more sophisticated methods available for years, particularly useful for "rolling terrain". Quicksurf could do some fairly fancy "smooth surfaces" in addition to TINs and has been widely used in mining and for modelling mineral deposits for years. It couldn't do overhangs but I can get along without them. For most man made surfaces (roads, building sites, etc) TINs seem to work well and I'm not in any hurry to change.

(Note: Delauney is not the surface type. Rather, it is the means by which triangles are chosen to from a random set of points to form a TIN.)

Title: Well now that you mention it...

By: Blake Beard

Posted: March 11, 2010 8:28 PM

Good thoughts here. I was just getting to the point where I can get my job done without fighting the software and I was quite happy in my ignorant bliss. Now that you mention it I see the interface changing but the functioality hasn't improved much at all! Mr. Ward now I get it!

Title: Yes - we need a new method

By: Sue Albermarle

Posted: March 12, 2010 8:45 AM

I am surrounded by very bright people and they often remark how civil design hasnt progressed much. Perhaps this is why, the software makers are sitting on their butts writing code based on formulas handed to them. But whose formulas and how did their "research departments" vet them out & improve them? Clearly they haven't.

The author is correct, that is why we are all known as dinosaurs in the engineering world. We dont change until we are obsolete, like surveying is becoming as an industry. I understand we now have less than 25K licensed surveyors in America according to the CLSA. Thanks for "Keepin them honest"! SA

Title: Do we?

By: Matthew Anderson, PE

Posted: March 12, 2010 9:49 AM

Harry - Nice article, however, do we really need a new method? or do we need to increase the understanding of the methods currently in use? Your contour smoothing method obviously does not effect the underlying surface TIN. It begs the question of having adequate information to create a smooth surface.
I have always wished to don my virtual tool belt, apply the 3d googles, and grade my site with bulldovers, and a collection of manholes.

Title: Thank you all for your comments !

By: Harry O. Ward, PE

Posted: March 16, 2010 9:36 AM

Thank you all for your comments - Matthew is correct, most users only use a small handful of tools in the software because corporate training is woefully poor in our industry. Users rarely know what their software is capable of. Our training includes education and theory and these weaknesses/failings continually come up in discussion. I often wished that the manufacturers could hear what is said in training classes. Now they will.

I have also received several other emails not reflected here and some indicated that there are alternatives with better algorithms on the market however, they are stand-alone solutions that are not contained in integrated products.

But the point is that although lots of time is spent on user interface and making things look pretty, not a lot of actual progress has been made on the algorithms being used. I started with one of the most simple, DTM's. I will continue to identify other areas within our industry that are out of date or badly needing improvement.

H. Ward

Title: i wasnt going to comment but...

By: Bob Moorfield

Posted: March 22, 2010 5:05 PM

i wasnt going to comment but now i feel i have to. i have a scanning job where i must analyze the tunnel we just scanned. i just found out that the software i am using can not handle tunnels. i am now upset and YES we need new algorithms that work on all of the jobs we do in our business. Keep up the good writing.

Title: New algorithms

By: Sam the Scanner

Posted: March 25, 2010 11:53 AM

Yes we need new algorithms, that is the first time i ever used that term, BTW. Scanning has brought new challenges and i would like to think that we are using modern techniques, i guess we are not. We should be. Thanks or the article.

Title: Good article -

By: JT

Posted: March 29, 2010 12:59 PM

i just saw an article in another magazine talking about how to fix a DTM machine control model by adding contours to it. OMG, are these people in the dark ages? No, you are supposed to use breaklines, not contours! I have heard many people say "contours are output, not input". Keeping it real!

Title: More on methods

By: Chad Schaeding

Posted: March 30, 2010 9:02 AM

The Delaunay triangulation technique is still a useful method in neighborhood selection, even within the framework of geostatistical interpolation techniques such as krieging (or other weighted distance operators). I have used it in previous work to get first, second, and third neighboring triangles to identify what neighboring samples to included in the krieging estimate. All surface models interpolate elevations between sample locations, even TIN or nearest neighbor, and the real question to ask is how many and what neighboring sampled locations do I use to estimate the unknown elevation?

Krieging is ideal for this because it can be flexible enough to include parameters that allow the operator to custom fit variograms and neighborhoods for the exact local terrain. Additionally it declusters samples and allows for variance estimates, thereby provide QA/QC of the volume estimates if variance propagation is added. The tradeoff is the increased complexity increase the cost of a krieging interpolated surface compared to that of a TIN. For unskilled users, a TIN might be preferred due to its simplicity and that no parameter (number of neighbors, variogram type and sub parameters) selection is required.

Another question to think about is: are different techniques preferred for topographic mapping and volume estimation of existing topography versus designed DTMs? What method is used when real-time updating right on-board the machine guided equipment? Is a hybrid approach appropriate, such as one that uses a smooth estimator in natural terrain and uses planar estimates on hard surfaces?

Our company performs estimates using several techniques to determine a probable range that the exact volume would be within, with high and low estimates. We are now just starting to work with excavators in the area of machine guidance and looking to expand our services into this market. Does anyone have any information about entering into the machine guidance data prep market?