No subject

[Hong Zhuang]

I am replying the following question received from email :

-----------------------------------------
< The field with which we are using GRASS is to analyze and digitize
< coastlines and then measure the amount of coastal erosion we can
< recognize from year to year (we were fortunate enough to acquire
< air photos of this region spanning over 25 years, in a rapidly changing
< coastline, with rates of recession exceeding 2-4 meters in spots).
<   In any case, the digitization of the coastline can be a tedious process,
< which to me seems like it could be automated, through image enhancement
< to pick out the cliff face of the shoreline. The shoreline is reasonably
< pronounced by dark hues on one side, and the other side is complete beach
< (light hues). This differentiation between the two contrasts should be easily
< distinguishable to software packages equipped to handle it.
<    I am hoping that any of the grass users have found a way to achieve this,
< perhaps through GRASS, or a secondary software package. I would be eager
< and appreciative to hear any suggestions you might have on the matter.
< All answers can be mailed to :
<      franklin at ug.cs.dal.ca
<    thank you in advance for your time.

< Steve

< p.s: the image I"m dealing with is much like the following:

< XXXXXXXXXXXXXXXXiiiii
< XXXXXXXXXXXXXiiiiiiii
< XXXXXXXXXXXXXXiiiiiii
< XXXXXXXXXXXXXXXiiiiiii
< XXXXXXXXXXXXXXXiiiiiii
< XXXXXXXXXXXXXXiiiiiiiii
< XXXXXXXXXXXXiiiiiiiiiii
< XXXXXXXXXXXXiiiiiiiiii
< XXXXXXXXXXXXiiiiiiiiiiii
< XXXXXXXXXXXXXiiiiiiiii
< XXXXXXXXXXXXXXiiiiiiiii
< XXXXXXXXXXXXXXXiiiiiiii
              ^
< (dark tundra) | (sandy beaches)
<               |
<               ---coastline

< Other regions are distinctly light hued on either side of the coastline.

< I humbly ask for your forgiveness if I have approached the forum with an
< inappropriate question. However, I would like to avoid re-inventing the
< wheel, and am hoping someone has worked in a similar direction as that which
< I have depicted.
-----------------------------------------

	As I understand, you project is belong to boundary finding problem.  
You need to find the boundary between X (dark tundra) and i (sandy beaches).  
Any boundary finder software can do this job.  In GRASS, we have a tool i.zc 
which is a kind of boundary finder and will find your coastlines.  GRASS tool 
i.zc first filters the image using Gaussian evolution.  And then conducts 
secondary derivatives using Laplasian operation. Finally it finds zero points 
as the boundary.  Therefore i.zc as be called as zero-crossing boundary finder. 
This is a sophisticated image processing tool and is existing in the GRASS 
package.  You may enjoy using it.
	Although the usage of any computer tool like the GRASS tool i.zc is
direct forward and simply, the interpretation of the results from the computer
is not so sinple some time.  For the tool i.zc, we have been studying the
errors of the results for quite a long time recently.  After you run i.zc, you
may find that some time the results are simply garbag and entirely un-trustable.
But most time the results are believable and only with some amount of deviation.
I presented a talk in the ASPRS/ACSM/RT 92 convention in Washington DC in 
August about how large of the deviation of computer calculated boundary from
the real boundary.  There are four situations accouding to different values
of Gaussian size you choose and coastline curvature. They are deviation-free
region, exaggeration region, fractal region and wipe-out region.  We have fond 
the criteria of how to identify the situations.  If you are interested to read 
the paper and have difficulty to find the proceedings,  I would be very glad to 
send you a copy. 
	Please try i.zc and let me know the results if you like.  I am very 
interested in this kind of problems.	Thank you. 

	Hong C. Zhuang
	EC Division, CERL
	P.O.Box 4005
	Champaign, IL 61820
	(217)352-6511 ext.588