如何检测表开始在iTextSharp的？(How to Detect table start in

2019-08-19 01:52发布

站内文章 / 移动开发
83 0 0

我想将PDF转换为csv文件。 PDF文件在表格格式与第一行作为报头数据。 我已经达到的水平，我可以从细胞中提取文本，比较文本的基线表和检测换行，但我需要比较表格边框检测表的开始。 我不知道如何检测和PDF比较线。 谁能帮我？

谢谢！！！

Answer 1:

正如你已经看到了（希望），PDF文件没有表的概念，只是文本放置在他们周围绘制的具体地点和线路。 有文本和线条之间没有内在联系。 理解这一点非常重要。

知道了这一点，如果所有的细胞都具有足够的填充您可以查找足够大，如3个或更多空间的宽度字符之间的差距。 如果细胞没有足够的空间，这将很可能不幸打破。

您还可以看看在PDF的每一行，并尝试找出代表你的“表样”行。 见这个答案怎么走每个令牌的网页上看到什么正在绘制。



Answer 2:

我也寻找类似问题的答案，但不幸的是我没有找到一个，这样我做到了我自己。

像这样的PDF页面

将给输出

这里是GitHub的链接，我做了DOTNET控制台应用程序。 https://github.com/Justabhi96/Detect_And_Extract_Table_From_Pdf

此应用程序检测到PDF的特定页面的表格，并将它们打印在控制台上的表格格式。 这里是我用来做这个应用程序的代码。

首先，我把文本从PDF的使用延伸iTextSharp.text.pdf.parser.LocationTextExtractionStrategy类iTextSharp的一类其坐标一起。 该代码如下：

这是要去块存储与有坐标和文字类。 

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;
        }

    }
}

这是一个扩展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()));
        }
    }
}

这个类重写LocationTextExtractionStrategy类会在您每次提取使用PdfTextExtractor.GetTextFromPage（）方法从PDF页面大块时间叫的RenderText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，为了检测表的开始和结束时，您可以使用从PDF中提取的块的坐标。 就像如果具体线路是不是有表则会有当前块的在右边没有跳转协调和下一个块的左侧坐标。 但线具有表将是那些具有坐标至少3个点的跳跃。

像其表将有大块这样的坐标系：

当前块的右坐标 - > 12.75pts
一个块的左COORDS - > 20.30pts

因此，进一步的你可以用这样的逻辑来检测PDF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;
        }
    }
}

您可以拨打上面的类的getLineText（）方法和运行下面的循环看到控制台上的表结构的输出。 

    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();


文章来源: How to Detect table start in itextSharp?
标签： pdf csv itextsharp pdf-reader pdf-parsing
举报

收藏的人(0)

Ta的文章 更多文章
登录 后发表评论
0条评论
还没有人评论过~