Maxluli/OCR_Decode
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 2 Wiki Activity

Cleaned Window.cs

Browse Source
This commit is contained in:
Rohmer Maxime
2023-04-25 15:16:08 +02:00
parent 7cb27ca69c
commit 32b20ff317
4 changed files with 110 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files
OCR_Decode/OcrImage.cs
+11 -11
View File
@@ -54,7 +54,7 @@ namespace OCR_Decode
/// Enhances the image depending on wich type of window the image comes from
/// </summary>
/// <param name="type">The type of the window. Depending on it different enhancing features will be applied</param>
/// <returns></returns>
/// <returns>The enhanced Bitmap</returns>
public Bitmap Enhance(WindowType type = WindowType.Text)
{
Bitmap outputBitmap = (Bitmap)InputBitmap.Clone();
@@ -128,7 +128,7 @@ namespace OCR_Decode
/// Method that convert a colored RGB bitmap into a GrayScale image
/// </summary>
/// <param name="inputBitmap">The Bitmap you want to convert</param>
/// <returns></returns>
/// <returns>The bitmap in grayscale</returns>
public static Bitmap Grayscale(Bitmap inputBitmap)
{
Rectangle rect = new Rectangle(0, 0, inputBitmap.Width, inputBitmap.Height);
@@ -165,7 +165,7 @@ namespace OCR_Decode
/// </summary>
/// <param name="inputBitmap">the bitmap you want to convert to binary colors</param>
/// <param name="threshold">The floor at wich the color is considered as white or black</param>
/// <returns></returns>
/// <returns>The binarised bitmap</returns>
public static Bitmap Tresholding(Bitmap inputBitmap, int threshold)
{
Rectangle rect = new Rectangle(0, 0, inputBitmap.Width, inputBitmap.Height);
@@ -203,7 +203,7 @@ namespace OCR_Decode
/// Method that removes the pixels that are flagged as background
/// </summary>
/// <param name="inputBitmap">The bitmap you want to remove the background from</param>
/// <returns></returns>
/// <returns>The Bitmap without the background</returns>
public static Bitmap RemoveBG(Bitmap inputBitmap)
{
Rectangle rect = new Rectangle(0, 0, inputBitmap.Width, inputBitmap.Height);
@@ -237,7 +237,7 @@ namespace OCR_Decode
/// Method that removes all the useless things from the image and returns hopefully only the numbers
/// </summary>
/// <param name="inputBitmap">The bitmap you want to remove useless things from (Expects a cropped part of the TyreWindow)</param>
/// <returns></returns>
/// <returns>The bitmap with (hopefully) only the digits</returns>
public unsafe static Bitmap RemoveUseless(Bitmap inputBitmap)
{
//Note you can use something else than a cropped tyre window but I would recommend checking the code first to see if it fits your intended use
@@ -336,7 +336,7 @@ namespace OCR_Decode
/// Recovers the average colors from the Image. NOTE : It wont take in account colors that are lower than the background
/// </summary>
/// <param name="inputBitmap">The bitmap you want to get the average color from</param>
/// <returns></returns>
/// <returns>The average color of the bitmap</returns>
public static Color GetAvgColorFromBitmap(Bitmap inputBitmap)
{
Rectangle rect = new Rectangle(0, 0, inputBitmap.Width, inputBitmap.Height);
@@ -383,7 +383,7 @@ namespace OCR_Decode
/// This method simply inverts all the colors in a Bitmap
/// </summary>
/// <param name="inputBitmap">the bitmap you want to invert the colors from</param>
/// <returns></returns>
/// <returns>The bitmap with inverted colors</returns>
public static Bitmap InvertColors(Bitmap inputBitmap)
{
Rectangle rect = new Rectangle(0, 0, inputBitmap.Width, inputBitmap.Height);
@@ -415,7 +415,7 @@ namespace OCR_Decode
/// </summary>
/// <param name="inputBitmap">The bitmap you want to resize</param>
/// <param name="resizeFactor">The factor of resizing you want to use. I recommend using even numbers</param>
/// <returns></returns>
/// <returns>The bitmap witht the new size</returns>
public static Bitmap Resize(Bitmap inputBitmap, int resizeFactor)
{
var resultBitmap = new Bitmap(inputBitmap.Width * resizeFactor, inputBitmap.Height * resizeFactor);
@@ -432,7 +432,7 @@ namespace OCR_Decode
/// method that Highlights the countours of a Bitmap
/// </summary>
/// <param name="inputBitmap">The bitmap you want to highlight the countours of</param>
/// <returns></returns>
/// <returns>The bitmap with countours highlighted</returns>
public static Bitmap HighlightContours(Bitmap inputBitmap)
{
Bitmap outputBitmap = new Bitmap(inputBitmap.Width, inputBitmap.Height);
@@ -475,7 +475,7 @@ namespace OCR_Decode
/// </summary>
/// <param name="inputBitmap">The bitmap you want to erode</param>
/// <param name="kernelSize">The amount of Erosion you want (be carefull its expensive on ressources)</param>
/// <returns></returns>
/// <returns>The Bitmap with the eroded contents</returns>
public static Bitmap Erode(Bitmap inputBitmap, int kernelSize)
{
Bitmap outputBitmap = new Bitmap(inputBitmap.Width, inputBitmap.Height);
@@ -534,7 +534,7 @@ namespace OCR_Decode
/// </summary>
/// <param name="inputBitmap">The bitmap you want to use dilatation on</param>
/// <param name="kernelSize">The amount of dilatation you want (be carefull its expensive on ressources)</param>
/// <returns></returns>
/// <returns>The Bitmap after Dilatation</returns>
public static Bitmap Dilatation(Bitmap inputBitmap, int kernelSize)
{
Bitmap outputBitmap = new Bitmap(inputBitmap.Width, inputBitmap.Height);
OCR_Decode/Reader.cs
+3 -3
View File
@@ -200,7 +200,7 @@ namespace OCR_Decode
/// Method that calls all the zones and windows to get the content they can find on the image to display them
/// </summary>
/// <param name="idImage">The id of the image we are working with</param>
/// <returns></returns>
/// <returns>a string representation of all the returns</returns>
public async Task<string> Decode(int idImage)
{
string result = "";
@@ -236,7 +236,7 @@ namespace OCR_Decode
/// Method that can be used to convert an amount of miliseconds into a more readable human form
/// </summary>
/// <param name="amountOfMs">The given amount of miliseconds ton convert</param>
/// <returns></returns>
/// <returns>A human readable string that represents the ms</returns>
public static string ConvertMsToTime(int amountOfMs)
{
//Convert.ToInt32 would round upand I dont want that
@@ -250,7 +250,7 @@ namespace OCR_Decode
/// Old method that can draw on an image where the windows and zones are created. mostly used for debugging
/// </summary>
/// <param name="idImage">the #id of the image we are working with</param>
/// <returns></returns>
/// <returns>the drawed bitmap</returns>
public Bitmap Draw(int idImage)
{
Bitmap result;
OCR_Decode/Window.cs
+88 -44
View File
@@ -1,4 +1,10 @@
using System;
/// Author : Maxime Rohmer
/// Date : 25/04/2023
/// File : Window.cs
/// Brief : Parent for futur windows and framework for the childrens
/// Version : 0.1
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
@@ -20,13 +26,14 @@ namespace OCR_Decode
public Rectangle Bounds { get => _bounds; private set => _bounds = value; }
public Bitmap Image { get => _image; set => _image = value; }
public string Name { get => _name; protected set => _name = value; }
//This will have to be changed if you want to make it run on your machine
public static DirectoryInfo TESS_DATA_FOLDER = new DirectoryInfo(@"C:\Users\Moi\Pictures\SeleniumScreens\TessData");
public Bitmap WindowImage
{
get
{
//This little trickery lets you have the image that the window sees
Bitmap sample = new Bitmap(Bounds.Width, Bounds.Height);
Graphics g = Graphics.FromImage(sample);
g.DrawImage(Image, new Rectangle(0, 0, sample.Width, sample.Height), Bounds, GraphicsUnit.Pixel);
@@ -40,36 +47,61 @@ namespace OCR_Decode
Engine = new TesseractEngine(TESS_DATA_FOLDER.FullName, "eng", EngineMode.Default);
Engine.DefaultPageSegMode = PageSegMode.SingleLine;
}
/// <summary>
/// Method that will have to be used by the childrens to let the model make them decode the images they have
/// </summary>
/// <returns>Returns an object because we dont know what kind of return it will be</returns>
public virtual async Task<Object> DecodePng()
{
return "NaN";
}
public virtual async Task<Object> DecodePng(List<string> drivers)
/// <summary>
/// Method that will have to be used by the childrens to let the model make them decode the images they have
/// </summary>
/// <param name="driverList">This is a list of the different possible drivers in the race. It should not be too big but NEVER be too short</param>
/// <returns>Returns an object because we dont know what kind of return it will be</returns>
public virtual async Task<Object> DecodePng(List<string> driverList)
{
return "NaN";
}
public static byte[] ImageToByte(Image img)
/// <summary>
/// This converts an image into a byte[]. It can be usefull when doing unsafe stuff. Use at your own risks
/// </summary>
/// <param name="inputImage">The image you want to convert</param>
/// <returns>A byte array containing the image informations</returns>
public static byte[] ImageToByte(Image inputImage)
{
using (var stream = new MemoryStream())
{
img.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
inputImage.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
return stream.ToArray();
}
}
public static async Task<int> GetTimeFromPng(Bitmap wImage,OcrImage.WindowType type, TesseractEngine Engine)
/// <summary>
/// This method is used to recover a time from a PNG using Tesseract OCR
/// </summary>
/// <param name="windowImage">The image where the text is</param>
/// <param name="windowType">The type of window it is</param>
/// <param name="Engine">The Tesseract Engine</param>
/// <returns>The time in milliseconds</returns>
public static async Task<int> GetTimeFromPng(Bitmap windowImage,OcrImage.WindowType windowType, TesseractEngine Engine)
{
//Kind of a big method but it has a lot of error handling and has to work with three special cases
string rawResult = "";
int result = 0;
switch (type)
switch (windowType)
{
case OcrImage.WindowType.Sector:
//The usual sector is in this form : 33.456
Engine.SetVariable("tessedit_char_whitelist", "0123456789.");
break;
case OcrImage.WindowType.LapTime:
//The usual Lap time is in this form : 1:45:345
Engine.SetVariable("tessedit_char_whitelist", "0123456789.:");
break;
case OcrImage.WindowType.Gap:
//The usual Gap is in this form : + 34.567
Engine.SetVariable("tessedit_char_whitelist", "0123456789.+");
break;
default:
@@ -78,7 +110,7 @@ namespace OCR_Decode
}
Bitmap enhancedImage = new OcrImage(wImage).Enhance(type);
Bitmap enhancedImage = new OcrImage(windowImage).Enhance(windowType);
var tessImage = Pix.LoadFromMemory(ImageToByte(enhancedImage));
@@ -105,8 +137,8 @@ namespace OCR_Decode
List<string> rawNumbers;
//In the gaps we can find '+' but we dont care about it its redondant
if(type == OcrImage.WindowType.Gap)
//In the gaps we can find '+' but we dont care about it its redondant a driver will never be - something
if(windowType == OcrImage.WindowType.Gap)
rawResult = Regex.Replace(rawResult, "[^0-9.:]", "");
//Splits into minuts seconds miliseconds
@@ -125,8 +157,6 @@ namespace OCR_Decode
{
//ss:ms
result = (Convert.ToInt32(rawNumbers[0]) * 1000) + Convert.ToInt32(rawNumbers[1]);
//1239
//931
if (result > 999999)
{
@@ -160,17 +190,24 @@ namespace OCR_Decode
}
}
}
//enhancedImage.Save(Reader.DEBUG_DUMP_FOLDER + Regex.Replace(rawResult, "[^0-9A-Za-z]", "") + ".png");
page.Dispose();
return result;
}
public static async Task<string> GetStringFromPng(Bitmap image, TesseractEngine Engine, string allowedChars = "",OcrImage.WindowType windowType = OcrImage.WindowType.Text)
/// <summary>
/// Method that recovers strings from an image using Tesseract OCR
/// </summary>
/// <param name="WindowImage">The image of the window that contains text</param>
/// <param name="Engine">The Tesseract engine</param>
/// <param name="allowedChars">The list of allowed chars</param>
/// <param name="windowType">The type of window the text is on. Depending on the context the OCR will behave differently</param>
/// <returns>the string it found</returns>
public static async Task<string> GetStringFromPng(Bitmap WindowImage, TesseractEngine Engine, string allowedChars = "",OcrImage.WindowType windowType = OcrImage.WindowType.Text)
{
string result = "";
Engine.SetVariable("tessedit_char_whitelist", allowedChars);
Bitmap rawData = image;
Bitmap rawData = WindowImage;
Bitmap enhancedImage = new OcrImage(rawData).Enhance(windowType);
Page page = Engine.Process(enhancedImage);
@@ -182,35 +219,36 @@ namespace OCR_Decode
result += iter.GetText(PageIteratorLevel.Word);
} while (iter.Next(PageIteratorLevel.Word));
}
/*
if (allowedChars.Contains("S"))
{
enhancedImage.Save(Reader.DEBUG_DUMP_FOLDER + "Tyre" + result +".png");
}
else
{
enhancedImage.Save(Reader.DEBUG_DUMP_FOLDER + result +".png");
}
*/
page.Dispose();
return result;
}
protected Bitmap GetSmallBitmapFromBigOne(Bitmap bmp, Rectangle rectangle)
/// <summary>
/// Get a smaller image from a bigger one
/// </summary>
/// <param name="inputBitmap">The big bitmap you want to get a part of</param>
/// <param name="newBitmapDimensions">The dimensions of the new bitmap</param>
/// <returns>The little bitmap</returns>
protected Bitmap GetSmallBitmapFromBigOne(Bitmap inputBitmap, Rectangle newBitmapDimensions)
{
Bitmap sample = new Bitmap(rectangle.Width, rectangle.Height);
Bitmap sample = new Bitmap(newBitmapDimensions.Width, newBitmapDimensions.Height);
Graphics g = Graphics.FromImage(sample);
g.DrawImage(bmp, new Rectangle(0, 0, sample.Width, sample.Height), rectangle, GraphicsUnit.Pixel);
g.DrawImage(inputBitmap, new Rectangle(0, 0, sample.Width, sample.Height), newBitmapDimensions, GraphicsUnit.Pixel);
return sample;
}
protected static string FindClosestMatch(List<string> array, string target)
/// <summary>
/// Returns the closest string from a list of options
/// </summary>
/// <param name="options">an array of all the possibilities</param>
/// <param name="testString">the string you want to compare</param>
/// <returns>The closest option</returns>
protected static string FindClosestMatch(List<string> options, string testString)
{
var closestMatch = "";
var closestDistance = int.MaxValue;
foreach (var item in array)
foreach (var item in options)
{
var distance = LevenshteinDistance(item, target);
var distance = LevenshteinDistance(item, testString);
if (distance < closestDistance)
{
closestMatch = item;
@@ -220,39 +258,45 @@ namespace OCR_Decode
return closestMatch;
}
//This method has been generated with the help of ChatGPT
protected static int LevenshteinDistance(string s1, string s2)
/// <summary>
/// Method that computes a score of distance between two strings
/// </summary>
/// <param name="string1">The first string (order irrelevant)</param>
/// <param name="string2">The second string (order irrelevant)</param>
/// <returns>The levenshtein distance</returns>
protected static int LevenshteinDistance(string string1, string string2)
{
if (string.IsNullOrEmpty(s1))
if (string.IsNullOrEmpty(string1))
{
return string.IsNullOrEmpty(s2) ? 0 : s2.Length;
return string.IsNullOrEmpty(string2) ? 0 : string2.Length;
}
if (string.IsNullOrEmpty(s2))
if (string.IsNullOrEmpty(string2))
{
return string.IsNullOrEmpty(s1) ? 0 : s1.Length;
return string.IsNullOrEmpty(string1) ? 0 : string1.Length;
}
var d = new int[s1.Length + 1, s2.Length + 1];
for (var i = 0; i <= s1.Length; i++)
var d = new int[string1.Length + 1, string2.Length + 1];
for (var i = 0; i <= string1.Length; i++)
{
d[i, 0] = i;
}
for (var j = 0; j <= s2.Length; j++)
for (var j = 0; j <= string2.Length; j++)
{
d[0, j] = j;
}
for (var i = 1; i <= s1.Length; i++)
for (var i = 1; i <= string1.Length; i++)
{
for (var j = 1; j <= s2.Length; j++)
for (var j = 1; j <= string2.Length; j++)
{
var cost = (s1[i - 1] == s2[j - 1]) ? 0 : 1;
var cost = (string1[i - 1] == string2[j - 1]) ? 0 : 1;
d[i, j] = Math.Min(Math.Min(d[i - 1, j] + 1, d[i, j - 1] + 1), d[i - 1, j - 1] + cost);
}
}
return d[s1.Length, s2.Length];
return d[string1.Length, string2.Length];
}
}
}
ReadMe.md
+6
View File
@@ -1,3 +1,9 @@
# OCR decode
Project that takes config files in the JSON format and decodes screenshots from the F1TV data channel.
If you want to use it go change thoses variables :
Reader.cs "DEBUG_DUMP_FOLDER"
Window.cs "TESS_DATA_FOLDER"