Re: [sm3600-devel] Hacked solution for homerun.c

[Andy L. <sm3...@me...>]

On Fri, 2002-01-25 at 04:54, Marian Eichholz wrote:
> Andy Longton wrote:
> > 
> > Ramzy Darwish wrote:
> > >Hello again,
> > >I tried a really simple idea for getting homerun to work for me.
> > >
> > >All I changed was one line in sm3600homerun.c:
> > >if (bHolesOk && (lMedian>=CHASSIS_GRAY_LEVEL || lMedian >=
> > >BLACK_BED_LEVEL)){
> > >    printf("AT HOME POSITION\n");
> > >    return ltHome;
> > >  }
> > >
> > >I kind of understand what I did, and it works, not from sane, but with
> > >scantool.
> > >
> > >Please tell me if this is good, bad, dirty or whatever.
> 
> Sorry, Andy, I did not even have a *chance* yet to look at Your reports.
> I am terribly busy these days :-((( Hope, it's getting better soon...

No complaints here!  Maybe I should pay you more? :)  (I am getting
flack from the final owner, but have been able to hold them off for
now.  If you've ever worked Tech Support, think 'bi-polar customer from
hell'.)

Were the output results from tweaking BLACK_HOLE_GREY any help?

> > I applied it, and tried it out with
> > 
> >         ./scantool -d 4 2> output.txt
> > 
> > Here's the contents of output.txt;
> > ------------------------------------------------------
> > skipped 1 class/vendor specific interface descriptors
> > debug: i=532 c=107 #0=586 i=1744 c=99 #1=1794 i=2947 c=86 i=3034 c=8
> > i=3046 c=4 i=5300 c=0
> 
> Uh, here we have a clue: The circle at 3000 is ripped into pieces (586
> being ok, 1794 being ok, the third one is missing due to small diameter
> (86 is not ok).

I hate attachments, but I hope others on the list will indulge me this
time.  The scan.pnm was resaved to .png so it should be under 50K.

Looking at the scan, I see four defects (from left to right on the
image);

* Below the middle hole - a small rectangle defect on the metal edge.

* Right hand side of the last hole - a small piece of lint(?)

* 2~ inches to the right of the last hole - some faint metal defects; to
me they look like the start of small hole punches

* 3~ inches to the right of the last hole - same faint metal defects

Would you suspect any of these?  In a few moments, I'll give it a shake
again to see if I can move the lint(?) and report back if anything
changes.

> Either there is dirt on the line (perhaps You want to check the raw scan
> again) or You simply want to make the minimum hole with somewhat
> smaller.
> 
> 
> > 012333033333333343440444444444444014444444443433333333333222 - 90
> > debug: i=5300 c=0
> 
> looking good after all, although the image is a bit dark. With my
> scanner 5-7 in the bright middle is to be expected.
> 
> > 012233333333333344444444444444444444444444444434333333333322 - 98
> > debug: i=5300 c=0
> > 012233333333333444333444444444444444444443344343333333333322 - 96
> > debug: i=5300 c=0
> > 012233333333333333433444444444444433444444343433333333333322 - 95
> > debug: i=5300 c=0
> > 022223333333333334345444444444444344444444444433333333333322 - 95
> > debug: i=552 c=69 i=1753 c=80 i=2955 c=83 i=5300 c=0
> > 012233033333333333340444444444443044444344443334333333333322 - 90
> > debug: i=536 c=100 #0=586 i=1740 c=106 #1=1793 i=2942 c=108 #2=2996
> > 012233033333334333430444444444444004434444434433333333333322 - 89
> 
> This one should have been a correct homerun: 3 circles detected with
> 100/106/108 pixels diameter at 586/1793/2996.
> 
> Right?

Sounds about right.