OCR recognition PDF file

1 Existing resolution pdf Methods
Use org.apache.pdfbox Read pdf, Can only read pdf The text , Some papers are scanned into pdf The words will be confused , Some words are still displayed in the form of pictures , The content read is incomplete , Often you can't get the data you want .

2 OCR Character recognition
pdf Need to convert to picture , For identification , High recognition rate .

2.1 Call Baidu interface
advantage ： High recognition rate , Fast recognition

shortcoming ： Charge per time

2.2 Use open source tools to read pdf file
2.2.1 Download Kit
https://github.com/tesseract-ocr/tessdata download chi_sim.traineddata,chi_sim_vert.traineddata

2.2.2 Add dependency

<dependencies>
        <dependency>
            <groupId>net.sourceforge.tess4j</groupId>
            <artifactId>tess4j</artifactId>
            <version>4.4.0</version>
        </dependency>
 </dependencies>

2.2.3 Programming

import net.sourceforge.tess4j.Tesseract;
import net.sourceforge.tess4j.TesseractException;
public class TestTess {
    
         public static void main(String[] args) {
    
        //  Identify the path of the picture （ Change to your own picture path ）
       String path = "C:\\work\\notebook\\prototype\\target\\qq1.jpg";
       //  Language library location （ Change it to the path of your own language Library folder ）
      // String lagnguagePath = "C:\\work\\projects\\tess\\resources\\tessdata";
       File file = new File(path);
       ITesseract instance = new Tesseract();
       // Set the location of the training library 
       //instance.setDatapath(lagnguagePath);
       //chi_sim ： Simplified Chinese , eng  Select the language library according to the needs 
       instance.setLanguage("chi_sim");
       String result = null;
       try {
    
           long startTime = System.currentTimeMillis();
           result =  instance.doOCR(file);
           long endTime = System.currentTimeMillis();
           System.out.println("Time is：" + (endTime - startTime) + "  millisecond ");
       } catch (TesseractException e) {
    
           e.printStackTrace();
       }
       System.out.println("result: ");
       System.out.println(result);                 
         }
}

2.3 Read pdf Location specific data
2.3.1 Manual interception pdf Specifies the material of the rectangular area
Need front-end cooperation to make pages , Change to automatic acquisition , When adding document types in this way , I need to configure it .

2.3.2 Python The program obtains the rectangular coordinates of the image identification area in the whole document

import aircv
def matchImg(imgsrc, imgobj, confidence=0.2):
    """  Picture contrast recognition imgobj stay imgsrc Relative position on （ Batch identify the parts needed in the unified picture ） :param imgsrc:  Original image path (str) :param imgobj:  Image path to be found （ Templates ）(str) :param confidence:  Recognition （0<confidence<1.0） :return: None or dict({'confidence':  Similarity degree (float), 'rectangle':  Rectangular coordinates on the original picture (tuple), 'result':  Central coordinates (tuple)}) """
    imsrc = aircv.imread(imgsrc)
    imobj = aircv.imread(imgobj)
    match_result = aircv.find_template(imsrc, imobj,
                                    confidence)  # {'confidence': 0.5435812473297119, 'rectangle': ((394, 384), (394, 416), (450, 384), (450, 416)), 'result': (422.0, 400.0)}
    if match_result is not None:
        match_result['shape'] = (imsrc.shape[1], imsrc.shape[0])  # 0 For the high ,1 To be wide 
    return match_result
template = {
    'address':'dz.jpg','doc_num':'fw.jpg','doc_type':'fwlx.jpg','issue_date':'fwrq.jpg',
              'int_cls': 'splb.jpg','apply_num':'sqh.jpg','applyer':'sqr.jpg','content':'zw.jpg'}
for key, value in template.items():

    orig = matchImg("target/qq.jpg","target/"+value)

    rect = orig['rectangle']

    w = rect[3][0] - rect[0][0]

    h = rect[3][1] - rect[0][1]

    x = rect[0][0]

    y = rect[0][1]

    ret = [x,y,w,h]

    print(key,ret)

2.3.3 Java The program is based on the coordinates of the rectangular area , Get the specified location information

package odysssey.tess;
import java.io.File;
import java.io.IOException;
import java.util.List;
import javax.imageio.ImageIO;
import java.awt.Rectangle;
import net.sourceforge.tess4j.ITessAPI.TessPageIteratorLevel;
import net.sourceforge.tess4j.ITesseract;
import net.sourceforge.tess4j.Tesseract;
import net.sourceforge.tess4j.TesseractException;
public class TestTess {
    
         public static void main(String[] args) throws IOException {
    
        //  Identify the path of the picture （ Change to your own picture path ）
       String path = "C:\\work\\notebook\\prototype\\target\\qq.jpg";
       //  Language library location （ Change it to the path of your own language Library folder ）
      // String lagnguagePath = "C:\\work\\projects\\tess\\resources\\tessdata";
       File file = new File(path);
       ITesseract instance = new Tesseract();
       // Set the location of the training library 
       /* address [61, 312, 734, 82] doc_num [1002, 338, 527, 78] doc_type [425, 736, 801, 82] issue_date [64, 593, 495, 64] int_cls [115, 969, 346, 82] apply_num [676, 589, 388, 68] applyer [72, 450, 481, 68] content [107, 899, 1439, 70] */    
         int rects[][] = {
    {
    61, 312, 734, 82},

                                         {
    1002, 338, 527, 78},

                                            {
    425, 736, 801, 82},

                                            {
    64, 593, 495, 64},

                                            {
    115, 969, 346, 82},

                                            {
    676, 589, 388, 68},

                                            {
    72, 450, 481, 68},

                                            {
    107, 899, 1439, 70}};
         
       //chi_sim ： Simplified Chinese , eng  Select the language library according to the needs 

       instance.setLanguage("chi_sim");
       instance.setTessVariable("user_defined_dpi", "96");
       String result = null;
       try {
    
           long startTime = System.currentTimeMillis();
           //result = instance.doOCR(file);
           for(int i = 0 ;i < rects.length;i++){
    
               Rectangle rr = new Rectangle(rects[i][0],rects[i][1],rects[i][2],rects[i][3]);
                    result =instance.doOCR(file, rr);
               System.out.println(result);                 
           }
    /* List<Rectangle> resul=instance.getSegmentedRegions(ImageIO.read(file), TessPageIteratorLevel.RIL_SYMBOL); for (int i = 0; i < resul.size(); i++) { Rectangle rect = resul.get(i); System.out.println(String.format("Box[%d]: x=%d, y=%d, w=%d, h=%d", i, rect.x, rect.y, rect.width, rect.height)); } */     long endTime = System.currentTimeMillis();
           System.out.println("Time is：" + (endTime - startTime) + "  millisecond ");
       } catch (TesseractException e) {
    
           e.printStackTrace();
       }
    }
}