unit1.pas

首先综述了手写体汉字识别问题及其研究现状

unit Unit1;

interface

uses
  Windows, Messages, SysUtils, Variants, Classes, Graphics, Controls, Forms,
  Dialogs, StdCtrls, ExtCtrls, ComCtrls;

const
  iOcrTypeMultiLine = $0; //要识别为多行文字
  iOcrTypeOneLine = $10000; //仅识别为一行文字
  iOcrTypeOneCharacter = $20000; //仅识别为一个字

  InfoAssociate = 1; //使用联想词库
  InfoAutoRotateBeforeRecognize = 2; //自动倾斜更正
  InfoSameFont = 4; //图片中字体大小只有一种
  InfoReservePositionCannotRecognize = 8; //不能识别的文字显示为?
  InfoUniteSameXInOneLine = $10; //强制合并同一行中的X坐标相同的点阵(主要用于文字中有大量白斑或断线的情况,其它情况下一般不要设置此值)
  InfoRotate180 = $20; //尝试旋转180度来识别
  InfoRecognizeTable = $40; //是否分析表格
  
  iGrayAuto = $10000; //灰度转为单色时自动设置阀值

  HzDigit = $10; //数字
  HzEnglish = $20; //英文
  HzInterpunction = $40; //其它符号
  HzChinese = $80; //中文

  HzOnlyone = $100; //仅单字节,不允许双字节英文、数字

  HZAll = HZDigit or HZEnglish or HzInterpunction or HzChinese; //所有字
  HZEDI = HZDigit or HZEnglish or HzInterpunction; //所有数字、英文、符号
  HZEDIOnlyone = HZEDI or HzOnlyone; //所有数字、英文、符号(单字节,不允许双字节英文、数字)
  HZDI = HZDigit or HzInterpunction; //所有数字、符号

  MaxSuspect = 15;
type
  TResult = record
    Chinese: array[0..3] of char;
    Precision: word;
    CountSuspect: word;
    ChineseSuspect: array[0..MaxSuspect * 2 + 1] of char;
    PrecisionSuspect: array[0..MaxSuspect] of word;
    Rect: TRect;
  end;
  PResult = ^TResult;
  TOCRResult = packed record
    Count: integer;
    Result: array[0..1] of TResult;
  end;
  POCRResult = ^TOCRResult;
  TForm1 = class(TForm)
    PageControl1: TPageControl;
    TabSheet1: TTabSheet;
    Image1: TImage;
    Panel1: TPanel;
    Button1: TButton;
    TabSheet2: TTabSheet;
    Memo1: TMemo;
    Button2: TButton;
    Buttontif2bmp: TButton;
    ButtonScan: TButton;
    ButtonScanAll: TButton;
    Button3: TButton;
    Button4: TButton;
    OpenDialog1: TOpenDialog;
    Button5: TButton;
    procedure Button1Click(Sender: TObject);
    procedure FormCreate(Sender: TObject);
    procedure Button2Click(Sender: TObject);
    procedure Buttontif2bmpClick(Sender: TObject);
    procedure ButtonScanClick(Sender: TObject);
    procedure ButtonScanAllClick(Sender: TObject);
    procedure Button3Click(Sender: TObject);
    procedure Button4Click(Sender: TObject);
    procedure Button5Click(Sender: TObject);
  private
    { Private declarations }
  public
    { Public declarations }
  end;

function WOcrInit(iType, iInfo, iCharacterSet: integer; res1: integer): BOOL; stdcall; external 'OcrWencui.dll';
function WOcrStart(HBmp: HBitmap;
  rRect: PRect; iGray, res: integer; var pOCRResult: POCRResult; var TextResult: pchar; OutputFile: Pchar): BOOL; stdcall; external 'OcrWencui.dll';
function WOcrStartFile(Filename: pchar; iPage: integer;
  rRect: PRect; iGray, res: integer; var pOCRResult: POCRResult; var TextResult: pchar; OutputFile: Pchar): BOOL; stdcall; external 'OcrWencui.dll';
procedure WOcrStop; stdcall; external 'OcrWencui.dll';
function WConvertToBmp(FileSrc, BmpFileDst: pchar; iPage: integer): bool; stdcall; external 'OcrWencui.dll';
function WConvertColors(FileSrc, BmpFileDst: pchar; iPage, Colors, Value: integer): bool; stdcall; external 'OcrWencui.dll';
function WConvertKeepColor(FileSrc, BmpFileDst: pchar; iPage: integer; cColor: TColor; RedValue, GreenValue, BlueValue: byte): boolean; stdcall; external 'OcrWencui.dll';
function WScanerSelect(Scanname: pchar): bool; stdcall; external 'OcrWencui.dll';
function WScanParameter(Scanname: pchar; bInvert, bDouble: bool; ImageType, ImageRes, ImageBright, ImageContrast: integer; rect: PRect): bool; stdcall; external 'OcrWencui.dll'; //Rect的单位:毫米
function WScanStartOnePage(filename: pchar): integer; stdcall; external 'OcrWencui.dll';
function WScanStartMultiPage(bShowGUI: boolean; iFileType: integer; Filename: pchar; bOverwrite: bool; iMaxScanPageCount, iMultiFirstPage, iStep: integer): integer; stdcall; external 'OcrWencui.dll';

var
  Form1: TForm1;

implementation

{$R *.dfm}

procedure TForm1.Button1Click(Sender: TObject);
var
  p: POCRResult;
  i: integer;
  s: string;
  Ptext: pchar;
begin
//  if WOcrStartFile('test.bmp',0, nil, iGrayAuto, 0, p, Ptext,
//    pchar(extractfilepath(paramstr(0)) + '1.pdf')) then
  if WOcrStart(image1.Picture.Bitmap.Handle, nil, iGrayAuto, 0, p, Ptext,
    pchar(extractfilepath(paramstr(0)) + '1.rtf')) then
  begin
    //显示每个汉字
    memo1.Lines.Clear;
    for i := 0 to p^.Count - 1 do
      with p.Result[i] do
      begin
        setlength(s, CountSuspect * 2);
        move(ChineseSuspect[0], s[1], CountSuspect * 2);
        memo1.Lines.Add(format('%-4s  相似度:%3d%% Left:%-4d Right:%-4d Top:%-4d Bottom:%-4d 候选字:%s', [string(Chinese), Precision,
          Rect.left, Rect.right, Rect.top, Rect.bottom, s]));
      end;
      //显示全部结果
    memo1.Lines.Add(strpas(ptext));
    if PageControl1.ActivePageIndex <> 1 then
      PageControl1.ActivePageIndex := 1;
  end;
end;

procedure TForm1.FormCreate(Sender: TObject);
begin
  image1.Picture.LoadFromFile(extractfilepath(paramstr(0)) + 'test.bmp');
  if not WOcrInit(iOcrTypeMultiLine or HZAll, //多行文字,所有字符
    InfoAssociate or InfoRecognizeTable, //使用联想词库,分析表格
    0, //字符集
    0 //保留未用
    ) then
    showmessage('初始化dll出错，请检查使用许可是否过期或与作者联系.');
end;

procedure TForm1.Button2Click(Sender: TObject);
var
  p: POCRResult;
  i: integer;
  s: string;
  r: trect;
  Ptext: pchar;
begin
  r.left := 156;
  r.Right := 507;
  r.Top := 40;
  r.Bottom := 100;
  if WOcrStart(image1.Picture.Bitmap.Handle, @r, iGrayAuto, 0, p, Ptext, '') then
  begin
    memo1.Lines.Clear;
    for i := 0 to p^.Count - 1 do
      with p.Result[i] do
      begin
        setlength(s, CountSuspect * 2);
        move(ChineseSuspect[0], s[1], CountSuspect * 2);
        memo1.Lines.Add(format('%-4s  相似度:%3d%% Left:%-4d Right:%-4d Top:%-4d Bottom:%-4d 候选字:%s', [string(Chinese), Precision,
          Rect.left, Rect.right, Rect.top, Rect.bottom, s]));
      end;
    memo1.Lines.Add(strpas(ptext));
    if PageControl1.ActivePageIndex <> 1 then
      PageControl1.ActivePageIndex := 1;
  end;
end;

procedure TForm1.Buttontif2bmpClick(Sender: TObject);
var
  path: string;
begin
  path := extractfilepath(paramstr(0));
  if WConvertToBmp(pchar(path + 'test.tif'), pchar(path + 'bak.bmp'), 0) then
    showmessage('test.tif已转为bak.bmp');
end;

procedure TForm1.ButtonScanClick(Sender: TObject);
var
  scaner: array[0..1024] of char;
  filename: string;
begin
  filename := extractfilepath(paramstr(0)) + 'bak.bmp';
  if WScanerSelect(@Scaner[0]) then
    if WScanParameter(@scaner, false, false, 0, 200, -127, 0, nil) then
      if WScanStartOnePage(pchar(filename)) <> 0 then
      begin
        showmessage('扫描完成,保存为' + filename);
        image1.Picture.LoadFromFile(filename);
      end;
end;

procedure TForm1.ButtonScanAllClick(Sender: TObject);
var
  scaner: array[0..1024] of char;
  filename: string;
  iPage: integer;
begin
  filename := extractfilepath(paramstr(0)) + 'bak';
  if WScanerSelect(@Scaner[0]) then
    if WScanParameter(@scaner, false, false, 0, 60, 0, 0, nil) then
    begin
      iPage := WScanStartMultiPage(false, 0, pchar(filename), true, maxint, 1, 2);
      if iPage <> 0 then
      begin
        showmessagefmt('扫描完成%d页', [iPage]);
        image1.Picture.LoadFromFile(filename + '001.bmp');
      end;
    end;
end;

procedure TForm1.Button3Click(Sender: TObject);
begin
  if WConvertColors('GLITE.BMP', 'bak.bmp', 0, 1, 128) then
    showmessage('ok');
end;

procedure TForm1.Button4Click(Sender: TObject);
var
  color: TColor;
begin
  color := $0404FC; //04蓝 04绿 红FC
  if WConvertKeepColor('GLITE.BMP', 'bak.bmp', 0, color, 20, 20, 20) then
    showmessage('ok');
end;

procedure TForm1.Button5Click(Sender: TObject);
begin
  if OpenDialog1.Execute then
    Image1.Picture.LoadFromFile(OpenDialog1.FileName); 
end;

end.