Retrieving and analyzing the results

Position along the X- and Y-axis.

The height and width of the character.

The match score of the character.

The spacing of the character.

The validity of the character (based on the character's acceptance level).

The ASCII version of the character.

The contents of the string.

The angle of the string.

The length of the string.

The match score of the string.

The validity of the string (based on the string's acceptance level).

All the characters read, even if they are on different lines in the target image.

The number of characters read in total.

The match score for the entire text.

The threshold value used to binarize the target image. Note that this is only available when using an M_GENERAL OCR font context, multiple strings, and/or the string has a few degrees of rotation.

If characters are vertically overlapping (or touching).

Create a child buffer that contains the characters below the point where they overlap. This effectively removes the overlapping portion of the characters.

The more that characters overlap vertically, the less chance for a valid result.

To read a target image containing multiple lines of text, MIL OCR requires that there is enough space between lines. If lines are too close together (vertically), the robustness of the search will suffer.

Use MocrControl() with M_TOUCHING_CHAR to read horizontally touching characters.

If the string or strings have non-uniform inter-character spacing.

Read/verify the string with an M_GENERAL OCR font context type and use MocrControl() with M_TARGET_CHAR_SPACING set to M_ANY.

This setting is best suited to read non-uniformly spaced characters.

Create a child buffer to contain the most extreme spacing examples for best results.

If the string or strings in the target image are greatly disparate.

Create a child buffer for each different region in position, angle, and/or length. Read each region individually for best results.

The more similar the strings in angle, position, and length, the faster and more robust the search.

If the spacing, length or angle differ, use different MIL OCR font contexts for each child buffer, with their operational and processing controls configured appropriately for each situation. For more information, see the Defining the target strings section earlier in this chapter.

If the target image contains multiple lines with different lengths.

Create a child buffer that contains the area with the shorter strings and read the shorter strings, using a different OCR font context, into a different OCR result buffer.

Strings of similar length are easiest to locate.

Make sure that the target image has a consistent inter-line spacing, and that all lines start at a similar location along the X-axis, for best results.

If the target image contains blanks.

Verify that your string length is correct.

Your string length determines the number of characters sought.

If the problem persists and the ability to read blank spaces is enabled (use MocrControl() with M_BLANK_CHARACTERS) remember to include them in your string length (use MocrControl() with M_STRING_CHAR_NUMBER).

If the target image contains a lot of non-character blobs.

Try to improve your target image before trying to read/verify again. For more information, see the Defining the target strings section earlier in this chapter.

An image that does not include blobs of equal intensity as the foreground will return best results.

Use MocrControl() with M_MORPHOLOGIC_FILTERING to internally enhance the contrast of the image.

If the target image contains a string longer or shorter than what was expected.

MIL OCR will try to return the number of characters that you set as existing in the target image using MocrControl() with M_STRING_CHAR_NUMBER.

If your string length is set to a value larger than the actual length of the string in the target image, and if using M_GENERAL OCR font context, MIL will return the best matches found and stop searching after it has found the specified number of characters. If using an M_CONSTRAINED OCR font context, the string length determines the exact number of characters found, even if this means returning characters with low match scores.

If the target string is at an angle.

Verify that the principal axis of the entire text is equal to the expected angle (nominal angle), specified using MocrControl() with M_STRING_ANGLE.

A string at an angle not equal to 0 takes longer to find.

Create a target image with a string angle of 0. If the string is properly read, your original application did not follow the expected use of string angle.