Shape Detection

The 'Sample Images' property allows you to provide one or more images of the shape(s) you want to detect in your documents. Each sample image should be a clear, representative example of the target shape, such as a stamp, seal, or logo.

You can add multiple sample images to detect several different shapes in a single pass. Each detected region will be associated with the sample image that produced the match.

Example: Add a sample image of a notary seal to detect its presence on scanned forms.

The 'Shape Name' property specifies the name that will be assigned to detected shapes. This name is used in Layout Data, classification features, and diagnostic output.

If left blank, the name of the matching sample image will be used for each result.

Example: Set to CompanyLogo to label all detected instances of your company logo.

Can be one of the following values:

The 'Proximity Measure' property determines how the similarity between the sample image and the document image is calculated. Choose from several methods, such as Sum of Absolute Differences (SAD), Cross-Correlation, or Sum of Squared Distances (SSD).

• 'SAD' is fast and effective for most shapes.
• 'CrossCorr' is more robust to lighting and contrast variations.
• 'SSD' is sensitive to outliers and may be useful for certain applications.

Tip: Try different methods if you are not getting reliable matches with the default.

The 'Background Differencing' property adjusts confidence values so that 0% represents a match against a blank region of the document. This is especially useful when shapes contain mostly whitespace, as it prevents blank areas from producing high confidence scores.

When enabled, confidence values are scaled according to the color balance of the sample image.

Example: Enable this property when detecting seals or logos with large white areas to avoid false positives on blank regions.

The 'Minimum Confidence' property sets the threshold for reporting a detected shape. Only matches with a confidence score greater than or equal to this value will be included in the results.

Specify a value between 0.0 (accept all matches) and 1.0 (only perfect matches).

Example: Set to 0.8 to require at least 80% similarity for a match to be reported.

The 'Maximum Angle' property defines the largest rotation (in degrees) that will be tested when searching for shapes. The system will test all angles from -Maximum Angle to +Maximum Angle in increments of 'Angle Step'.

Example: Set to 25 to detect shapes rotated up to 25° in either direction.

The 'Angle Step' property specifies the increment (in degrees) between each angle tested when searching for rotated shapes. Smaller values increase detection accuracy for rotated shapes but may increase processing time.

Example: With a 'Maximum Angle' of 25 and an 'Angle Step' of 5, the system will test -25°, -20°, -15°, -10°, -5°, 0°, 5°, 10°, 15°, 20°, 25°.

The 'Minimum Scale' property sets the smallest scale at which the sample image will be tested against the document image. Use this to detect shapes that may appear smaller than the sample image.

Specify a value between 10% (10% of original size) and 100% (original size).

Example: Set to 50% to detect shapes at half their original size or larger.

The 'Maximum Scale' property sets the largest scale at which the sample image will be tested against the document image. Use this to detect shapes that may appear larger than the sample image.

Specify a value between 100% (original size) and 400% (400% of original size).

Example: Set to 200% to detect shapes up to twice their original size.

The 'Scale Step' property specifies the increment between each scale factor tested when searching for resized shapes. Smaller values increase detection accuracy for shapes at varying sizes but may increase processing time.

Example: Set to 5% to test every 5% between the minimum and maximum scale.

The 'Complexity' property displays the total number of shape comparisons that will be performed, based on the number of sample images, scale factors, and angles to be tested. This helps you estimate the processing time required for your configuration.

Tip: Reduce the number of samples, scales, or angles to improve performance if needed.

Can be one of the following values:

The 'Processing Resolution' property determines the resolution (in DPI) at which shape detection is performed. This setting directly affects the speed, memory usage, and accuracy of the detection process.

• Lower resolutions (such as Dpi50) process images faster and use less memory, but may miss small or fine details. This is suitable for detecting large shapes or when processing high-resolution scans where performance is a concern.
• Higher resolutions (such as Dpi100 or FullResolution) provide greater accuracy for detecting small shapes or fine details, but require more processing time and memory.

Common values:

• Dpi100: Good balance for most forms and moderate-sized shapes.
• Dpi50: Fastest, suitable for large shapes or high-resolution images.

Guidance:

• For large stamps, seals, or logos on 300 DPI scans, Dpi50 is often sufficient and much faster.
• For small logos or fine details, or when working with low-resolution images, use Dpi100 or FullResolution.
• If you are troubleshooting missed detections, try increasing the resolution and review diagnostic images to see if the shape is being lost at lower resolutions.

Example: Set to 50 DPI for fast detection or 100 DPI for higher accuracy.

Can be one of the following types:

The 'Binarization' property allows you to specify how the input image is converted to black and white before shape detection. Proper binarization can improve detection accuracy, especially for color or grayscale images.

Configure the binarization method and parameters to suit your document type and image quality.

Example: Use adaptive binarization for forms with uneven lighting or background shading.

The 'Dilation Factor' property controls how many times the image is dilated before shape detection is performed. Dilation is a morphological image processing operation that expands the white (foreground) regions in a binary image, effectively thickening lines and connecting nearby features.

Applying dilation can help:

• Connect broken, faint, or dotted lines in degraded or low-quality images, making shapes more detectable.
• Fill small gaps or holes in the target shape, improving match confidence when the original is incomplete.

However, excessive dilation can:

• Cause nearby shapes or features to merge, potentially increasing false positives.
• Blur fine details, making it harder to distinguish between similar shapes.

Guidance:

• Use a value of 0 (no dilation) for clean, high-quality images where shapes are well-defined.
• Increase the value (e.g., 1 or 2) for noisy, degraded, or faint images where shapes are broken or incomplete.
• Start with a low value and increase only as needed, reviewing diagnostic images to ensure shapes are not being merged or lost.

Example:

• Set to 2 to apply two rounds of dilation before detection, which can help recover faint stamps or seals on old or low-contrast documents.

Best Practices:

• Always review the effect of dilation using diagnostic output to ensure it is improving detection without introducing new errors.
• If you observe false positives or merged shapes, reduce the dilation factor or improve image quality upstream.

The 'Region of Interest (inches)' property restricts shape detection to a specific area of the image. Use this to improve performance and reduce false positives when the shape is known to appear in a predictable location.

Specify the region using the zone editor or enter coordinates directly. If left empty, the entire image will be searched.

Example: Set the region to the lower-right corner of a form if a stamp always appears in that area.

The 'Command Type Name' property shows the user-friendly name of the IP Command as it appears in the Grooper UI. This helps users identify which command is being configured or executed, especially when working with IP Profiles containing multiple commands.

This property is read-only and is automatically set based on the command's type.

The 'Supported Pixel Formats' property lists the image pixel formats that the IP Command can process. If an image does not match one of these formats, the command will be skipped for that image.

• By default, all pixel formats are supported unless a derived command restricts this list.
• Example formats include Pixel8bppGrayscale, Pixel24bppBgr, and Pixel1bppIndexed.

Use this property to verify compatibility when building IP Profiles or troubleshooting why a command did not execute.