gusucode.com > vision工具箱matlab源码程序 > vision/+vision/+internal/+buildable/OCRBuildable.m
%#codegen
%#ok<*EMCA>
classdef OCRBuildable < coder.ExternalDependency
methods (Static)
function name = getDescriptiveName(~)
name = 'ocr';
end
function b = isSupportedContext(context)
b = context.isMatlabHostTarget();
end
function updateBuildInfo(buildInfo, context)
usesDefaultModname = true;
vision.internal.buildable.cvstBuildInfo(buildInfo, context, ...
'ocrutils', ...
{'use_tesseract','use_leptonica','use_cpp11compat'},...
usesDefaultModname);
% PackNGo Note: The language data file used by tesseract is not
% added to the buildInfo object because
% 1) tesseract requires the language data file to be saved in
% a folder named 'tessdata'
% 2) folders cannot be added to the buildInfo.
%
% Therefore, the packNGo zip file will not have the language
% data file included in it.
end
function [asciiText, ocrMetadata] = tesseract(tessOpts, I, hasROI, resetParameters)
vision.internal.buildable.errorIfInSimulink('ocr');
coder.internal.errorIf(hasROI,'vision:ocr:codegenROIUnsupported');
coder.cinclude('cvstCG_ocrutils.h');
sizeI = size(I);
height = int32(sizeI(1));
width = int32(sizeI(2));
assert(height > 0);
assert(width > 0);
utf8Text = coder.opaque('char *','NULL');
txtLength = int32(0);
textLayout = nullTerminateString(tessOpts.textLayout);
characterSet = nullTerminateString(tessOpts.characterSet);
tessdata = nullTerminateString(tessOpts.tessdata);
lang = nullTerminateString(tessOpts.lang);
tessAPIHandle = coder.opaque('void *','NULL');
if islogical(I)
%> Invoke tesseract and return UTF-8 formatted text.
txtLength = coder.ceval('tesseractRecognizeTextLogical', ...
coder.ref(tessAPIHandle),...
coder.ref(I),...
coder.ref(utf8Text),...
width,...
height,...
textLayout,...
coder.rref(characterSet),... % coder.rref usage forces characterSet var to be generated in code
tessdata,...
lang, ...
resetParameters);
elseif strcmpi(class(I),'uint8')
%> Invoke tesseract and return UTF-8 formatted text.
txtLength = coder.ceval('tesseractRecognizeTextUint8', ...
coder.ref(tessAPIHandle),...
coder.ref(I),...
coder.ref(utf8Text),...
width,...
height,...
textLayout,...
coder.rref(characterSet),... % coder.rref usage forces characterSet var to be generated in code
tessdata,...
lang, ...
resetParameters);
end
if txtLength >= 0
metadata = coder.opaque('void *','NULL');
numChars = coder.nullcopy(int32(0));
numWords = coder.nullcopy(int32(0));
numTextlines = coder.nullcopy(int32(0));
numParagraphs = coder.nullcopy(int32(0));
numBlocks = coder.nullcopy(int32(0));
coder.ceval('collectMetadata',...,
tessAPIHandle, ...,
coder.ref(metadata),...
coder.ref(numChars),...
coder.ref(numWords),...
coder.ref(numTextlines),...
coder.ref(numParagraphs),...
coder.ref(numBlocks));
charBBox = coder.nullcopy(zeros(numChars, 4));
charWordIndex = coder.nullcopy(zeros(numChars, 1,'int32'));
charConfidence = coder.nullcopy(zeros(numChars, 1,'single'));
wordBBox = coder.nullcopy(zeros(numWords, 4));
wordTextLineIndex = coder.nullcopy(zeros(numWords, 1,'int32'));
wordConfidence = coder.nullcopy(zeros(numWords, 1,'single'));
wordCharacterIndex = coder.nullcopy(zeros(numWords, 2,'int32'));
textlineBBox = coder.nullcopy(zeros(numTextlines, 4));
textlineParagraphIndex = coder.nullcopy(zeros(numTextlines, 1,'int32'));
textlineConfidence = coder.nullcopy(zeros(numTextlines, 1,'single'));
textlineCharacterIndex = coder.nullcopy(zeros(numTextlines, 2,'int32'));
paragraphBBox = coder.nullcopy(zeros(numParagraphs, 4));
paragraphBlockIndex = coder.nullcopy(zeros(numParagraphs, 1,'int32'));
paragraphConfidence = coder.nullcopy(zeros(numParagraphs, 1,'single'));
paragraphCharacterIndex = coder.nullcopy(zeros(numParagraphs, 2,'int32'));
blockBBox = coder.nullcopy(zeros(numBlocks, 4));
blockPageIndex = coder.nullcopy(zeros(numBlocks, 1,'int32'));
blockConfidence = coder.nullcopy(zeros(numBlocks, 1,'single'));
blockCharacterIndex = coder.nullcopy(zeros(numBlocks, 2,'int32'));
coder.ceval('copyMetadata',...
metadata, ...
coder.ref(charBBox), coder.ref(charWordIndex), coder.ref(charConfidence),...
coder.ref(wordBBox), coder.ref(wordTextLineIndex), coder.ref(wordConfidence), coder.ref(wordCharacterIndex),...
coder.ref(textlineBBox), coder.ref(textlineParagraphIndex), coder.ref(textlineConfidence), coder.ref(textlineCharacterIndex),...
coder.ref(paragraphBBox), coder.ref(paragraphBlockIndex), coder.ref(paragraphConfidence), coder.ref(paragraphCharacterIndex),...
coder.ref(blockBBox), coder.ref(blockPageIndex), coder.ref(blockConfidence), coder.ref(blockCharacterIndex));
ocrMetadata.CharacterBBox = charBBox;
ocrMetadata.CharacterWordIndex = charWordIndex;
ocrMetadata.CharacterConfidence = charConfidence;
ocrMetadata.WordBBox = wordBBox;
ocrMetadata.WordTextLineIndex = wordTextLineIndex;
ocrMetadata.WordConfidence = wordConfidence;
ocrMetadata.WordCharacterIndex = wordCharacterIndex;
ocrMetadata.TextLineBBox = textlineBBox;
ocrMetadata.TextLineParagraphIndex = textlineParagraphIndex;
ocrMetadata.TextLineConfidence = textlineConfidence;
ocrMetadata.TextLineCharacterIndex = textlineCharacterIndex;
ocrMetadata.ParagraphBBox = paragraphBBox;
ocrMetadata.ParagraphBlockIndex = paragraphBlockIndex;
ocrMetadata.ParagraphConfidence = paragraphConfidence;
ocrMetadata.ParagraphCharacterIndex = paragraphCharacterIndex;
ocrMetadata.BlockBBox = blockBBox;
ocrMetadata.BlockPageIndex = blockPageIndex;
ocrMetadata.BlockConfidence = blockConfidence;
ocrMetadata.BlockCharacterIndex = blockCharacterIndex;
tmp = coder.nullcopy(zeros(1,txtLength, 'uint8'));
%> Copy UTF-8 encoded text into temporary buffer and cleanup.
coder.ceval('copyTextAndCleanup', utf8Text,...
coder.ref(tmp),txtLength);
%> Convert UTF-8 text to ASCII text.
asciiText = char(utf8ToAscii(tmp));
if numel(asciiText) ~= numChars
% Text does not match what is in metadata. Get text
% from metadata instead to maintain consistency.
metadataText = coder.opaque('char *','NULL');
% get text from metadata
len = coder.nullcopy(zeros(1,'int32'));
len = coder.ceval('getTextFromMetadata', metadata, coder.ref(metadataText));
% copy text into output buffer
tmpText = coder.nullcopy(zeros(1,len, 'uint8'));
coder.ceval('copyTextAndCleanup', metadataText,...
coder.ref(tmpText),len);
%> Convert UTF-8 text to ASCII text.
ocrMetadata.Characters = char(utf8ToAscii(tmpText));
else
ocrMetadata.Characters = '';
end
coder.ceval('cleanupMetadata', metadata);
coder.ceval('cleanupTesseract', tessAPIHandle);
else
% An error occurred while running tesseract.
coder.internal.errorIf(txtLength == int32(-1),...
'vision:ocr:codegenInitFailure');
coder.internal.errorIf(txtLength == int32(-2), ...
'vision:ocr:codegenMemAllocFailure');
coder.internal.errorIf(txtLength == int32(-3), ...
'vision:ocr:codegenInternalError');
% codegen requires all execution paths to assign outputs
asciiText = char(zeros(1,0,'uint8'));
ocrMetadata.CharacterBBox = zeros(0, 4);
ocrMetadata.CharacterWordIndex = zeros(0, 1,'int32');
ocrMetadata.CharacterConfidence = zeros(0, 1,'single');
ocrMetadata.WordBBox = zeros(0, 4);
ocrMetadata.WordTextLineIndex = zeros(0, 1,'int32');
ocrMetadata.WordConfidence = zeros(0, 1,'single');
ocrMetadata.WordCharacterIndex = zeros(0, 2,'int32');
ocrMetadata.TextLineBBox = zeros(0, 4);
ocrMetadata.TextLineParagraphIndex = zeros(0, 1,'int32');
ocrMetadata.TextLineConfidence = zeros(0, 1,'single');
ocrMetadata.TextLineCharacterIndex = zeros(0, 2,'int32');
ocrMetadata.ParagraphBBox = zeros(0, 4);
ocrMetadata.ParagraphBlockIndex = zeros(0, 1,'int32');
ocrMetadata.ParagraphConfidence = zeros(0, 1,'single');
ocrMetadata.ParagraphCharacterIndex = zeros(0, 2,'int32');
ocrMetadata.BlockBBox = zeros(0, 4);
ocrMetadata.BlockPageIndex = zeros(0, 1,'int32');
ocrMetadata.BlockConfidence = zeros(0, 1,'single');
ocrMetadata.BlockCharacterIndex = zeros(0, 2,'int32');
ocrMetadata.Characters = '';
end
end
end
end
% -------------------------------------------------------------------------
% Convert UTF-8 encoded text into ASCII. Multibyte characters are truncated
% to char(127).
% -------------------------------------------------------------------------
function asciiText = utf8ToAscii(utf8Text)
coder.inline('never');
idx = findCharacters(utf8Text);
asciiText = utf8Text(idx);
asciiText(asciiText>uint8(127)) = uint8(127);
end
% -------------------------------------------------------------------------
% Returns indices to the start of characters in a UTF-8 encoded string,
% skipping over UTF-8 continuation bytes. The number of characters is also
% returned.
% -------------------------------------------------------------------------
function [idx, count, invalidIndex] = findCharacters(utf8String)
coder.inline('never');
bytesToProcess = length(utf8String);
idx = false(1,bytesToProcess);
invalidIndex = false(1,bytesToProcess);
i = 1; count = 0;
while bytesToProcess
count = count + 1;
idx(i) = true;
if isASCIICompatiable(utf8String(i))
i = i + 1;
bytesToProcess = bytesToProcess - 1;
elseif isUTF8ContinuationByte(utf8String(i))
invalidIndex(i) = true;
i = i + 1;
bytesToProcess = bytesToProcess - 1;
elseif isStartOfUTF8MultiByteCharacter(utf8String(i))
numBytes = numBytesInSequence(utf8String(i));
% skip over the multi-byte sequence
i = i + numBytes;
bytesToProcess = bytesToProcess - numBytes;
else
invalidIndex(i) = true;
i = i + 1;
bytesToProcess = bytesToProcess - 1;
end
end
end
% -------------------------------------------------------------------------
% Return true if the UTF-8 byte holds an ASCII compatible value.
% -------------------------------------------------------------------------
function tf = isASCIICompatiable(utf8Byte)
coder.inline('always');
tf = utf8Byte < uint8(128) ;
end
% -------------------------------------------------------------------------
% Return true if the UTF-8 byte is a continuation byte (10XX XXXX)
% -------------------------------------------------------------------------
function tf = isUTF8ContinuationByte(utf8Byte)
coder.inline('always');
tf = utf8Byte >= uint8(128) && utf8Byte < uint8(192);
end
% -------------------------------------------------------------------------
% Return true if byte is the start of a multi-byte UTF-8 character.
% -------------------------------------------------------------------------
function tf = isStartOfUTF8MultiByteCharacter(utf8Byte)
coder.inline('always');
tf = utf8Byte >= uint8(192) && utf8Byte <= uint8(252);
end
% -------------------------------------------------------------------------
% Return the number of bytes in a multi-byte UTF-8 character by inspecting
% the starting byte of the sequence.
% -------------------------------------------------------------------------
function n = numBytesInSequence(utf8Byte)
coder.inline('always');
n = 1;
while bitand(bitshift(uint8(utf8Byte),n),uint8(128))
n = n + 1; % tells us how many bytes are in the multi-byte char
end
assert(n<7); % invalid UTF-8 start byte
end
% -------------------------------------------------------------------------
% Null-terminate a string.
% -------------------------------------------------------------------------
function strout = nullTerminateString(strin)
coder.inline('always');
strout = [strin uint8(0)];
end