Form Recognition Settings
As of version 1.1 which was released in June 2010, FormReturn's recognition settings will detect either:
- Single checkboxes
- Multiple checkboxes with complete checkbox selection.
The new detection algorithm is very robust, which means heavy lines and borders added near the mark areas won't affect the detection of response data during the data capture process.
What Recognition Settings are Used for.
Recognition Settings are used by the form processor for determining which checkbox /es contain response marks. They are:
The default settings are normally the correct settings for detecting marks made in checkboxes. The settings won't have to be changed unless there is a problem with detection of marks in checkboxes.
A test preview of a scanned form will let you make sure the settings are correct for your forms.
Recognition Panel in the Recognition preview, where the recognition settings can be changed.
Mark Detection.
FormReturn detects marked checkboxes from scanned form images, by counting the white and black pixels contained inside the borders, as well as any stray marks that run outside the border and as far as the fragment edges.
View of a mark area in Test Preview
When the Mark Threshold
is set at 40, FormReturn will detect any checkbox that has a black
pixel count of over 40% (that means 40% more black pixels than the
white pixels)
If any of the marks weren't accurately detected, the Mark Threshold and Luminance settings can be changed. Any changes you make to the Recognition Settings will be
saved to the published form template and used in processing of the completed forms.
Increasing the Luminance
setting will make the image darker, for instance if pencil was used to mark the checkboxes, some of the grey color might be counted as white. Because FormReturn makes a comparison between the number of black to the number of white pixels, too much white
reduces the Mark threshold. Note: Making the luminance setting too high can
cause little black dots to appear on the scanned image which means checkboxes that weren't marked could be detected as marked.
Test your Recognition Settings
A Test preview
using a form image from your scanner, before Publishing a form template, will
determine if the detection of hand-marks will be accurate.
If you find that your scanned form images aren't clear enough to be detected, Scanning your form,
with increased resolution and / or contrast settings on your scanner,
will make a sharper scanned image so that the black parts of the image can be easily detected by the form processor.
Your form template should still be open in the Form Editor:
Click on the Recognition Preview Tab and click the Print button to print a form straight from the template.
Hand mark some of the checkboxes on the form using the same color pencil or pen that your respondents will be using, scan it and save to your computer, then open it by browsing for it in Test preview. This will test the scanned form image of the template before you publish it.
The image opened in test preview will determine the quality of images from your scanner and tests if
the default recognition settings will accurately detect hand marks in
checkboxes. If a pencil was used to mark the checkboxes, the Luminance setting may need to be changed if the hand-marks aren't detected properly.
You would increase the Mark Threshold setting, if the Mark Detection of black pixels isn't 40% or more greater than the counted white pixels in any checkbox.
Note: Any adjustments you make to the Recognition Settings, will be saved to the form template and used for processing of the returned forms.
Preview the Scanned Test Image:
- click the Recognition Preview Tab.
- Click the Test Preview Tab
- Browse for the scanned form image on your computer and open it.
Check Detect Marks in the Preview Check panel.
The checkboxes which were handmarked should be shown as marked by FormReturn. If the checkbox with the mark value of D was marked then the recognition test should read as Marked:(D)
Screenshot: a scanned form image opened in Test preview.
Mark Detection Details
If there is any error in detection of a marked checkbox, click on the mark area to open the Mark Detection Details dialog.
Mark Detection Details dialog.
FormReturn calculates the number of black pixels inside the borders of each checkbox ( this includes any marks that continue
outside the checkbox, but within the fragment padding).
Mark Area - You can use the Zoom feature to see how the black pixels would have been detected from the scanned image.
Silhouettes panel - In the image above, you can see that the marked checkbox (A) has 896 black pixels; 1001 pixels in total; 105 white pixels and a Threshold of 753%, which means the black pixel count is 753% more than the white pixel count.
Mark & Score Calculation - Any of these calculations can be changed before you publish the form template and the new settings will be saved to the publication. They will be used by the form processor for detecting response marks on the completed forms.
The Mark Threshold is 40% which is the default setting, and since the black pixel count is way more than that, there is no problem with the detection of the marked checkbox.
Mark Area Panel
Zooming in on the mark area will show you the quality of your scanned image and how much of the shading was determined as black by FormReturn.
Silhouettes Panel
- Pixels - The number of black pixels counted for each checkbox in the area.
- Enclosed - The total number of pixels counted (black and white) for each checkbox in the area.
- White - The number of white pixels counted for each checkbox in the area.
- Threshold - The percentage of black pixels more than white pixels counted for each checkbox in the area. Because the threshold is set at 40%, all checkboxes with 40% more black than white pixels will show as detected.
- Value - The mark value for each checkbox in the area.
The Mark and Score Calculation Panel
- Mark Threshold - setting that was used for the publication.
- Detected Marks - Mark value of the detected checkbox.
- Aggregate rule - set for the detected checkbox.
- Calculated Score - Score given to the detected checkbox using the Aggregate Rule.
- Field Name - Captured Data Fieldname or question name.
Luminance
The Luminance setting will only make a difference to the response mark detection in color scanned images. The setting (default 200) is used as the cut off point at which the pixels in checkboxes are determined as black or white. Some shades of grey, for instance in a pencil mark, would be determined as either black or white, depending on the clarity of the scanned image. If too much of the checkbox mark is determined as white, the mark won't be detected because the difference between black and white isn't high enough, for the black part of the mark to be detected.
Increasing the the luminance setting to around 210 will cause more black than white pixels to be detected, but if it is increased too much, it will cause black dots to appear in the image which is bad for detection.
Note: If the Luminance is set too high, it will cause black pixels to appear on the image where they shouldn't, leading to detection errors. Ideally the Luminance setting can be set anywhere between 100 and 220, but this will depend on the quality of the images your scanner produces.
Mark Threshold
The percentage that the Mark Threshold is set at is used by the form processor to detect the marked checkbox or checkboxes. The default threshold is 40% which means any checkbox with a count of 40% more black pixels than the white pixels, will be detected as a marked checkbox.
If an unmarked checkbox was detected as marked, the threshold is
probably not set high enough. If you can see that the threshold
percentage is less than the difference between the black and white
pixels, increasing the setting will fix the detection.
Click on a checkbox to open the Mark Detection Details dialog, to see the threshold percentages of the pixel count in each checkbox.
Fragment Padding
Select Detect Fragments from the Preview Check panel.
The Fragment padding is the white area between the checkbox borders and the red detected Fragment border.
Increasing the Fragment Padding setting, increases the amount of white area around the checkboxes.
The advantage of increasing the Fragment Padding, is if the scanned image quality isn't very good it improves detection because more of the Fragment area is captured in the data capture process.
The disadvantage of increasing the Fragment Padding is that it takes up more room on the form. If any other areas on the form touch or overlap the edges of the Fragment area, the risk is taken that the detection of the boxes could be broken.
Deskew Threshold:
It is highly recommended that the automatic Image Deskew box is left checked, unless you are completely sure your images were scanned in straight.
Keeping the Perform Deskew box checked, will automatically perform a Deskew. The setting determines the angle at which FormReturn will perform the deskew of a scanned image.
The Deskew Threshold setting indicates the angle at which an automatic deskew will be performed.
When the Deskew Threshold is set at the default of 1.05, FormReturn will automatically perform a Deskew if the Image is
more than 5 deg. skewed.
The Skew of the scanned Image is detected and is shown in degrees.
-0.03 deg skew means the form was scanned in on a slight angle which isn't enough for an auto deskew to be performed as it isn't enough to cause problems.