public DocumentCandidateNavigation(DocumentData pageOcr)
{
PageOcr = pageOcr;
Init();
}
public DocumentDataNavigation(DocumentData pageOcr)
{
PageOcr = pageOcr;
Init();
}
private static void LoadOcrData(DocumentData doc, string ocrFile, bool deskew)
{
var page = TPage.LoadFromPRD(ocrFile); //*.prd
// var pageNavigation = new TPageNavigation(page);
var data = doc.Words;
var lines = doc.Lines;
/* if (deskew)
* {
* foreach(FieldData field in doc.Fields)
* {
* field.OriginRectangle = field.Rectangle;
* TOCRRect origin = new TOCRRect((int)field.Rectangle.Left,(int)field.Rectangle.Top,(int)field.Rectangle.Width,(int)field.Rectangle.Height);
* var originSkew = page.Deskew( origin);
* field.Rectangle = new Rect(originSkew.Left, originSkew.Top, originSkew.Width, originSkew.Height);
* }
* }*/
List <TOCRRect> wordsRects = new List <TOCRRect>();
foreach (TLine line in page.Lines)
{
OcrLine ocrLine = new OcrLine();
ocrLine.Rectangle = new Rect(line.Rect.Left, line.Rect.Top, line.Rect.Width, line.Rect.Height);
foreach (TWord word in line.Words)
{
if (wordsRects.Any(x => x.Left == word.Rect.Left &&
x.Top == word.Rect.Top &&
x.Right == word.Rect.Right &&
x.Bottom == word.Rect.Bottom) == true)
{
continue;
}
else
{
wordsRects.Add(word.Rect);
}
// Split word
IList <MappedWord> spltWords = SplitWord(word)
.Select(x =>
{
var mapWord = new MappedWord()
{
Confidence = x.Confidance,
Contents = string.Concat(x.Chars.Select(a => a.CharData).ToArray())
};
TOCRRect deskewedRect = x.Rect;
if (deskew)
{
deskewedRect = page.Deskew(x.Rect);
}
mapWord.Rectangle = new Rect(deskewedRect.Left, deskewedRect.Top, deskewedRect.Width, deskewedRect.Height);
mapWord.OriginRectangle = new Rect(x.Rect.Left, x.Rect.Top, x.Rect.Width, x.Rect.Height);
return(mapWord);
})
.ToList();
// Add splited each words to the doc
for (int i = 0; i < spltWords.Count; i++)
{
MappedWord mapWord = spltWords[i];
if (i > 0)
{
spltWords[i - 1].SplitRight = mapWord;
}
if (i + 1 < spltWords.Count)
{
spltWords[i + 1].SplitLeft = mapWord;
}
doc.Words.Add(mapWord);
mapWord.Line = ocrLine;
ocrLine.Words.Add(mapWord);
}
}
if (ocrLine.Words.Any())
{
lines.Add(ocrLine);
}
}
int indexWord = 0;
int indexLine = 0;
foreach (var ocrLine in lines.OrderBy(a => a.Rectangle.Top).ToList())
{
ocrLine.ID = indexLine++;
}
data = data.OrderBy(a => a.Line.ID).ThenBy(a => a.Rectangle.X).ToList();
foreach (OcrWord ocrWord in data)
{
ocrWord.ID = indexWord++;
}
}
public static List <FieldClusterLine> CreateLines(DocumentData doc, fieldClusterModel cluster)
{
DocumentDataNavigation docNavigate = doc.WordsNavigator;
var wordList = cluster.Fields;
var wordHeight = CalcWordHeight(wordList);
List <FieldClusterLine> result = new List <FieldClusterLine>();
wordList = wordList.OrderBy(a => a.Rectangle.Top).ToList();
List <MappedWord> newWordList = new List <MappedWord>(wordList);
Dictionary <MappedWord, List <MappedWord> > wordLines = new Dictionary <MappedWord, List <MappedWord> >();
foreach (var word in newWordList)
{
var inTheSameLine = docNavigate.GetWords(new System.Windows.Rect(cluster.Area.Left, word.Rectangle.Top, cluster.Area.Width, word.Rectangle.Height))
.Where(x => x.Cluster.ID == word.Cluster.ID && x.Clusterline == null)
.Where(x => x == word || x.Rectangle.Top + (x.Rectangle.Height / 2) * 1.2 < word.Rectangle.Bottom)
.ToList();
if (inTheSameLine.Count == 0)
{
throw new Exception("A) No words in word line");
}
wordLines.Add(word, inTheSameLine);
}
int index = 0;
while (newWordList.Count > 0)
{
MappedWord WordTop = newWordList.FirstOrDefault(x => x.Rectangle.Height < wordHeight * 2);
if (WordTop == null)
{
WordTop = newWordList.First();
}
var baseWords = wordLines[WordTop].Where(x => x.Clusterline == null).ToList();
if (baseWords.Count == 0)
{
// Bad word
throw new Exception("B) No words in word line");
}
FieldClusterLine FieldLine = new FieldClusterLine();
FieldLine.ID = index++;
FieldLine.Fields = wordLines.Where(x => baseWords.Contains(x.Key))
.SelectMany(x => x.Value)
.Where(x => x.Clusterline == null)
.Distinct()
.OrderBy(x => x.Rectangle.Left)
.ToList();
newWordList.RemoveAll(a => FieldLine.Fields.Contains(a));
//FieldLine.Fields = newWordList.Where(a => WordTop.Line.Words.Contains(a)).ToList().OrderBy(b => b.Rectangle.X).ToList();
// FieldLine.Fields = newWordList.Where(a => (a.Rectangle.Top <= (WordTop.Rectangle.Top + (WordTop.Rectangle.Height * PrecentAsLine)))).ToList().OrderBy(b=>b.Rectangle.X).ToList();
FieldLine.Fields.ForEach(a => a.Clusterline = FieldLine);
result.Add(FieldLine);
}
return(result);
}
public static List <fieldClusterModel> ClusterListFromDoc(DocumentData doc)
{
List <fieldClusterModel> result = new List <fieldClusterModel>();
doc.Words.ToList().ForEach(a => result.Add(new fieldClusterModel((MappedWord)a)));
bool change = false;
do
{
change = false;
for (int i = 0; i < result.Count(); i++)
{
var fieldCluste1 = result.ElementAt(i);
if (fieldCluste1.isEmpty == false)
{
for (int j = 0; j < result.Count(); j++)
{
if (i == j)
{
continue;
}
var fieldCluste2 = result.ElementAt(j);
if (fieldCluste2.isEmpty == false)
{
if (fieldCluste2.AddIfIntersect(fieldCluste1))
{
fieldCluste1.Clear();
change = true;
break;
}
}
}
}
}
result.RemoveAll(a => a.isEmpty);
} while (change == true);
change = false;
do
{
change = false;
for (int i = 0; i < result.Count(); i++)
{
var fieldCluste1 = result.ElementAt(i);
if (fieldCluste1.isEmpty == false)
{
for (int j = 0; j < result.Count(); j++)
{
if (i == j)
{
continue;
}
var fieldCluste2 = result.ElementAt(j);
if (fieldCluste2.isEmpty == false)
{
if (fieldCluste2.AddIfInside(fieldCluste1))
{
fieldCluste1.Clear();
change = true;
break;
}
}
}
}
}
result.RemoveAll(a => a.isEmpty);
} while (change == true);
int index = 0;
result.ForEach(a => a.ID = index++);
foreach (MappedWord word in doc.Words)
{
word.Cluster = result.Where(a => a.Fields.Contains(word)).FirstOrDefault();
}
result.ForEach(a => a.lines = LineEngine.CreateLines(doc, a));
return(result);
}
