我正在尝试将 pdf 转换为 csv 文件。pdf文件具有表格格式的数据，第一行作为标题。我已经达到了可以从单元格中提取文本、比较表格中文本的基线并检测换行符的水平，但我需要比较表格边框以检测表格的开头。我不知道如何检测和比较 PDF 中的行。谁能帮我？
谢谢！！！
问问题
6170 次
2 回答
6
正如您所看到的（希望如此），PDF 没有表格的概念，只有放置在特定位置的文本和围绕它们绘制的线条。文本和行之间没有内在联系。理解这一点非常重要。
知道了这一点，如果所有单元格都有足够的填充，您可以寻找足够大的字符之间的间隙，例如 3 个或更多空格的宽度。如果单元格没有足够的间距，不幸的是这可能会破裂。
您还可以查看 PDF 中的每一行，并尝试找出代表“表格”行的内容。有关如何遍历页面上的每个标记以查看正在绘制的内容，请参阅此答案。
于 2013-04-02T15:32:19.353 回答
5
我也在寻找类似问题的答案，但不幸的是我没有找到，所以我自己做了。
像这样的PDF页面
将输出为
这是我制作的 dotnet 控制台应用程序的 github 链接。 https://github.com/Justabhi96/Detect_And_Extract_Table_From_Pdf
此应用程序检测 PDF 特定页面中的表格，并在控制台上以表格格式打印它们。这是我用来制作这个应用程序的代码。
首先，我使用扩展iTextSharp 的 iTextSharp.text.pdf.parser.LocationTextExtractionStrategy类的类从 PDF 中取出文本及其坐标。代码如下：
这是要存储带有坐标和文本的块的类。
using System;
using System.Collections.Generic;
using System.Linq;
using System.Web;
namespace itextPdfTextCoordinates
{
public class RectAndText
{
public iTextSharp.text.Rectangle Rect;
public String Text;
public RectAndText(iTextSharp.text.Rectangle rect, String text)
{
this.Rect = rect;
this.Text = text;
}
}
}
这是扩展LocationTextExtractionStrategy类的类。
using iTextSharp.text.pdf.parser;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Web;
namespace itextPdfTextCoordinates
{
public class MyLocationTextExtractionStrategy : LocationTextExtractionStrategy
{
public List<RectAndText> myPoints = new List<RectAndText>();
//Automatically called for each chunk of text in the PDF
public override void RenderText(TextRenderInfo renderInfo)
{
base.RenderText(renderInfo);
//Get the bounding box for the chunk of text
var bottomLeft = renderInfo.GetDescentLine().GetStartPoint();
var topRight = renderInfo.GetAscentLine().GetEndPoint();
//Create a rectangle from it
var rect = new iTextSharp.text.Rectangle(
bottomLeft[Vector.I1],
bottomLeft[Vector.I2],
topRight[Vector.I1],
topRight[Vector.I2]
);
//Add this to our main collection
this.myPoints.Add(new RectAndText(rect, renderInfo.GetText()));
}
}
}
此类覆盖 LocationTextExtractionStrategy 类的RenderText方法，每次使用PdfTextExtractor.GetTextFromPage()方法从 PDF 页面提取块时都会调用该方法。
using itextPdfTextCoordinates;
using iTextSharp.text.pdf;
//Create an instance of our strategy
var t = new MyLocationTextExtractionStrategy();
var path = "F:\\sample-data.pdf";
//Parse page 1 of the document above
using (var r = new PdfReader(path))
{
for (var i = 1; i <= r.NumberOfPages; i++)
{
// Calling this function adds all the chunks with their coordinates to the
// 'myPoints' variable of 'MyLocationTextExtractionStrategy' Class
var ex = iTextSharp.text.pdf.parser.PdfTextExtractor.GetTextFromPage(r, i, t);
}
}
//Here you can loop over the chunks of PDF
foreach(chunk in t.myPoints){
Console.WriteLine("character {0} is at {1}*{2}",i.Text,i.Rect.Left,i.Rect.Top);
}
现在，为了检测表格的开始和结束，您可以使用从 PDF 中提取的块的坐标。 就像如果特定行没有表格，那么当前块的右坐标和下一个块的左坐标将没有跳转。但是有表格的线条将有至少 3 个点的坐标跳跃。
像具有表的行将具有块的坐标，如下所示：
当前块的
右坐标 -> 12.75pts 下一个块的左坐标 -> 20.30pts
因此，您可以进一步使用此逻辑来检测 PDF 中的表格。代码如下：
using itextPdfTextCoordinates;
using iTextSharp.text.pdf;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace ConsoleApp1
{
class LineUsingCoordinates
{
public static List<List<string>> getLineText(string path, int page, float[] coord)
{
//Create an instance of our strategy
var t = new MyLocationTextExtractionStrategy();
//Parse page 1 of the document above
using (var r = new PdfReader(path))
{
// Calling this function adds all the chunks with their coordinates to the
// 'myPoints' variable of 'MyLocationTextExtractionStrategy' Class
var ex = iTextSharp.text.pdf.parser.PdfTextExtractor.GetTextFromPage(r, page, t);
}
// List of columns in one line
List<string> lineWord = new List<string>();
// temporary list for working around appending the <List<List<string>>
List<string> tempWord;
// List of rows. rows are list of string
List<List<string>> lineText = new List<List<string>>();
// List consisting list of chunks related to each line
List<List<RectAndText>> lineChunksList = new List<List<RectAndText>>();
//List consisting the chunks for whole page;
List<RectAndText> chunksList;
// List consisting the list of Bottom coord of the lines present in the page
List<float> bottomPointList = new List<float>();
//Getting List of Coordinates of Lines in the page no matter it's a table or not
foreach (var i in t.myPoints)
{
Console.WriteLine("character {0} is at {1}*{2}", i.Text, i.Rect.Left, i.Rect.Top);
// If the coords passed to the function is not null then process the part in the
// given coords of the page otherwise process the whole page
if (coord != null)
{
if (i.Rect.Left >= coord[0] &&
i.Rect.Bottom >= coord[1] &&
i.Rect.Right <= coord[2] &&
i.Rect.Top <= coord[3])
{
float bottom = i.Rect.Bottom;
if (bottomPointList.Count == 0)
{
bottomPointList.Add(bottom);
}
else if (Math.Abs(bottomPointList.Last() - bottom) > 3)
{
bottomPointList.Add(bottom);
}
}
}
// else process the whole page
else
{
float bottom = i.Rect.Bottom;
if (bottomPointList.Count == 0)
{
bottomPointList.Add(bottom);
}
else if (Math.Abs(bottomPointList.Last() - bottom) > 3)
{
bottomPointList.Add(bottom);
}
}
}
// Sometimes the above List will be having some elements which are from the same line but are
// having different coordinates due to some characters like " ",".",etc.
// And these coordinates will be having the difference of at most 4 points between
// their bottom coordinates.
//so to remove those elements we create two new lists which we need to remove from the original list
//This list will be having the elements which are having different but a little difference in coordinates
List<float> removeList = new List<float>();
// This list is having the elements which are having the same coordinates
List<float> sameList = new List<float>();
// Here we are adding the elements in those two lists to remove the elements
// from the original list later
for (var i = 0; i < bottomPointList.Count; i++)
{
var basePoint = bottomPointList[i];
for (var j = i+1; j < bottomPointList.Count; j++)
{
var comparePoint = bottomPointList[j];
//here we are getting the elements with same coordinates
if (Math.Abs(comparePoint - basePoint) == 0)
{
sameList.Add(comparePoint);
}
// here ae are getting the elements which are having different but the diference
// of less than 4 points
else if (Math.Abs(comparePoint - basePoint) < 4)
{
removeList.Add(comparePoint);
}
}
}
// Here we are removing the matching elements of remove list from the original list
bottomPointList = bottomPointList.Where(item => !removeList.Contains(item)).ToList();
//Here we are removing the first matching element of same list from the original list
foreach (var r in sameList)
{
bottomPointList.Remove(r);
}
// Here we are getting the characters of the same line in a List 'chunkList'.
foreach (var bottomPoint in bottomPointList)
{
chunksList = new List<RectAndText>();
for (int i = 0; i < t.myPoints.Count; i++)
{
// If the character is having same bottom coord then add it to chunkList
if (bottomPoint == t.myPoints[i].Rect.Bottom)
{
chunksList.Add(t.myPoints[i]);
}
// If character is having a difference of less than 3 in the bottom coord then also
// add it to chunkList because the coord of the next line will differ at least 10 points
// from the coord of current line
else if (Math.Abs(t.myPoints[i].Rect.Bottom - bottomPoint) < 3)
{
chunksList.Add(t.myPoints[i]);
}
}
// Here we are adding the chunkList related to each line
lineChunksList.Add(chunksList);
}
bool sameLine = false;
//Here we are looping through the lines consisting the chunks related to each line
foreach(var linechunk in lineChunksList)
{
var text = "";
// Here we are looping through the chunks of the specific line to put the texts
// that are having a cord jump in their left coordinates.
// because only the line having table will be having the coord jumps in their
// left coord not the line having texts
for (var i = 0; i< linechunk.Count-1; i++)
{
// If the coord is having a jump of less than 3 points then it will be in the same
// column otherwise the next chunk belongs to different column
if (Math.Abs(linechunk[i].Rect.Right - linechunk[i + 1].Rect.Left) < 3)
{
if (i == linechunk.Count - 2)
{
text += linechunk[i].Text + linechunk[i+1].Text ;
}
else
{
text += linechunk[i].Text;
}
}
else
{
if (i == linechunk.Count - 2)
{
// add the text to the column and set the value of next column to ""
text += linechunk[i].Text;
// this is the list of columns in other word its the row
lineWord.Add(text);
text = "";
text += linechunk[i + 1].Text;
lineWord.Add(text);
text = "";
}
else
{
text += linechunk[i].Text;
lineWord.Add(text);
text = "";
}
}
}
if(text.Trim() != "")
{
lineWord.Add(text);
}
// creating a temporary list of strings for the List<List<string>> manipulation
tempWord = new List<string>();
tempWord.AddRange(lineWord);
// "lineText" is the type of List<List<string>>
// this is our list of rows. and rows are List of strings
// here we are adding the row to the list of rows
lineText.Add(tempWord);
lineWord.Clear();
}
return lineText;
}
}
}
您可以调用上述类的getLineText()方法并运行以下循环以在控制台上查看表结构中的输出。
var testFile = "F:\\sample-data.pdf";
float[] limitCoordinates = { 52, 671, 357, 728 };//{LowerLeftX,LowerLeftY,UpperRightX,UpperRightY}
// This line gives the lists of rows consisting of one or more columns
//if you pass the third parameter as null the it returns the content for whole page
// but if you pass the coordinates then it returns the content for that coords only
var lineText = LineUsingCoordinates.getLineText(testFile, 1, null);
//var lineText = LineUsingCoordinates.getLineText(testFile, 1, limitCoordinates);
// For detecting the table we are using the fact that the 'lineText' item which length is
// less than two is surely not the part of the table and the item which is having more than
// 2 elements is the part of table
foreach (var row in lineText)
{
if (row.Count > 1)
{
for (var col = 0; col < row.Count; col++)
{
string trimmedValue = row[col].Trim();
if (trimmedValue != "")
{
Console.Write("|" + trimmedValue + "|");
}
}
Console.WriteLine("");
}
}
Console.ReadLine();
于 2018-09-15T08:34:08.807 回答