Issue 1130 in zxing: UNEXPECTED SCAN RESULTS WHILE USING LIBRARY

108 views
Skip to first unread message

zx...@googlecode.com

unread,
Jan 20, 2012, 1:43:11 AM1/20/12
to zx...@googlegroups.com
Status: New
Owner: ----

New issue 1130 by soorajna...@gmail.com: UNEXPECTED SCAN RESULTS WHILE
USING LIBRARY
http://code.google.com/p/zxing/issues/detail?id=1130

Hi,


This is Sooraj from Kerala, India. Couple of days back, I came across this
problem, which I can say is definitely a library based issue.

STEPS TO REPRODUCE THE ISSUE
____________________________

1. Install any app using zxing library for QR code detection on an android
device (v2.2, v2.3.4)
2. Print out a QR code with some text written just above it.
3. Scan the code in an inclined angle, with both the text and QR code
inside 'viewfinder' rectangle.
4. Adjust the device in such a way that both the text above and QR code
gets 'yellow-dots' above it
5. Check the result obtained (which definitely is an unexpected one)
As you can see from the attached files, I've some text written over QR
code, when I tried to capture the QR code in an inclined direction, so that
both the text and QR are inside the viewfinder rectangle, the reading gave
me unexpected result.


DESCRIPTION
___________

As you can see from the attached files, I've some text written over QR
code, when I tried to capture the QR code in an inclined direction, so that
both the text and QR are inside the viewfinder rectangle, the reading gave
me unexpected result.

But, in normal case, if I am scanning in a vertical direction(i.e. from
above the QR code), I am getting expected results. That you can find from
the second image.

I request someone to kindly help me with this issue. The issue that I found
is with the Barcode Scanner
application, and i guess this issue is there with every device. But just
fyi, I tried only on these devices,

Motorola Fire XT (android v2.3.4),
Samsung Galaxy Tab (android v2.2),
Samsung Galaxy S (android v2.2),

Any help would be well appreciated.


Thanks,

Sooraj

Attachments:
ERROR_IMAGE.png 21.2 KB
SUCCESS_IMAGE.png 21.4 KB

zx...@googlecode.com

unread,
Jan 20, 2012, 3:16:13 AM1/20/12
to zx...@googlegroups.com
Updates:
Status: NotABug

Comment #1 on issue 1130 by srowen: UNEXPECTED SCAN RESULTS WHILE USING
LIBRARY
http://code.google.com/p/zxing/issues/detail?id=1130

This is just a false positive, which can happen sometimes. It's not a bug.

zx...@googlecode.com

unread,
Jan 20, 2012, 4:07:46 AM1/20/12
to zx...@googlegroups.com

Comment #2 on issue 1130 by soorajna...@gmail.com: UNEXPECTED SCAN RESULTS

Just out of curiosity, can you please explain what this "false positive"
means...

I am experiencing this every time I follow the steps. Also, what you can
see from those images are, in the erroneous image output, the text is taken
in as a barcode.

There could be more users who might have experienced this flaw. Any
suggestions from you?

zx...@googlecode.com

unread,
Jan 20, 2012, 4:20:55 AM1/20/12
to zx...@googlegroups.com

Comment #3 on issue 1130 by srowen: UNEXPECTED SCAN RESULTS WHILE USING
LIBRARY
http://code.google.com/p/zxing/issues/detail?id=1130

That text happens to look like a barcode. It will happen some times, but is
quite rare. You are welcome to suggest a change? but remember that if
you 'fix' this you will cause real barcodes to not scan.

zx...@googlecode.com

unread,
Jan 20, 2012, 4:54:08 AM1/20/12
to zx...@googlegroups.com

Comment #4 on issue 1130 by soorajna...@gmail.com: UNEXPECTED SCAN RESULTS

Thanks for all your replies sire.

QRcoder1

unread,
Jan 20, 2012, 4:12:36 PM1/20/12
to zxing
Hello Sooraj

If the moderators may permit me to explain what is actually
hapenning. Thank You

I have been in the barcoding industry for many years and have seen
this problem before.

And this is what causes these errors.

**** Quiet zone too small **** move your text further away from the
code.
Ex: Some specialized barcoding software will permit you to print data
on any side of a QR code
if you do not set a " per specification " quiet zone you can and will
augment the chance of having
inapropriate decoding.

**** Improper scan angle **** scan your code at an appropriate
angle
CCD capture is the cause, since we are taking an image and converting
the pixelized data and try to interpret
code type algorithms, having additional digitized pixel in the field
and in the correct patterns can cause un wanted "Garbage" decodes.

Its simple optical physics, have a look at your pen or take a photo
of an object at an angle and the infield view of the object will be
correct,
but in the far field the bars or segments of a code will be smaller,
and can fall within a distinctive pattern that the scanner will "
correct " and perform the decode.

Wich falls within the realm of error correction.

**** Improper error correction **** augment value of error
correction

Scanning codes with low values or no error correction opens a door to
misinterpretation of data ( as in your example, correct me if i am
wrong ..i havent seen your files)
Augment the error correction of that barcode.

If you correct even just the quiet zone and the error correction you
can keep the human readable fairly close to the code.

But remember, and this is basic " go by the specification of the
barcode " and you shouldnt have these problems.

Trust me on this , i am certain it is not the scan " engine at fault "
but more of the above mentionned problems.

I have worked in conjuction with many manufacturers of barcoding
equipment including, laser scanners , ccd scanners and even
holographic scanners.
And i have seen to many examples of bad bar-coding, one company i have
worked for we also did barcode certification and i have seen many
weird situations.

A very good example of bad printing would be a QR code on a pen , the
curved surface and the fact the read angles are offset the image can
be corrected by the scanner and decoded ...but its not the actual
encoded code but a corrected interpretation.

Hope i dont loose to many of you on this one.....

Regards François Trudeau QRcoder1(dot)com



On Jan 20, 4:54 am, zx...@googlecode.com wrote:
> Comment #4 on issue 1130 by soorajna...@gmail.com: UNEXPECTED SCAN RESULTS
> WHILE USING LIBRARYhttp://code.google.com/p/zxing/issues/detail?id=1130

Sean Owen

unread,
Jan 20, 2012, 6:30:04 PM1/20/12
to zx...@googlegroups.com
That's all pretty good advice, but in this case it's not the scanning of the QR code that is at issue, but a false positive in scanning some other piece of the image.
It ought to scan the QR code just fine, and does. It's that a false positive jumps in before in some instances.

QRcoder1

unread,
Jan 20, 2012, 7:42:32 PM1/20/12
to zxing
Hello Sean

That is what i am mentionning in my post. Just in other terms.

{ Its simple optical physics, have a look at your pen or take a
photo
{ of an object at an angle and the infield view of the object will
be correct,
{ but in the far field the bars or segments of a code will be
smaller,
{ and can fall within a distinctive pattern that the

{ scanner will " correct " and perform the decode.

But the decoded barcode is actually garbage data.

Could you provide the 2 images from Sooraj please , i would like to
see how they are encoded, and how close the human readable is from the
code.

Is the scan engine used in zxing project your own source or it is a
standard CCD decode module from manufacturig engineers.


Thanks François Trudeau

Sean Owen

unread,
Jan 21, 2012, 6:42:08 AM1/21/12
to zx...@googlegroups.com
If you're referring to perspective distortion, there's none in this case. See the sample images in the issue.
There is no correction for perspective distortion going on anyway. There is also not a 1D barcode to begin with; it is not that some segments are distorted and misread.
I don't know what CCD decode module you're referring to; this is a question about this project.
Reply all
Reply to author
Forward
0 new messages