简单的文字识别

main.py

@@ -1,6 +1,6 @@
 import pinPicture
-from PIL import Image
 import tkinter as tk
+import numpy as np
 
 import pyautogui
 import keyboard
@@ -14,9 +14,6 @@ def capture_screen():
     root.mainloop()
 
 
-
-# 监听特定快捷键，这里假设是Ctrl+Alt+S
 keyboard.add_hotkey('F7', capture_screen)
 
-# 保持程序运行，持续监听快捷键
+keyboard.wait()
-keyboard.wait()

pinPicture.py

@@ -2,13 +2,15 @@ import io
 import math
 import tkinter as tk
 import win32clipboard
-from PIL import Image, ImageTk, ImageChops, ImageDraw, ImageGrab
+from PIL import Image, ImageTk, ImageDraw
-import threading
+import numpy as np
+import easyocr
 
 
 class PinPicture(tk.Frame):
     def __init__(self, image, master=None):
         super().__init__(master)
+        self.detectBtn = None
         self.pinBtn = None
         self.lineBtn = None
         self.arrowBtn = None
@@ -34,9 +36,10 @@ class PinPicture(tk.Frame):
         self.drawing = False
         self.drawMode = False
         self.color = "red"
-        self.image = image
         if image.mode != "RGBA":
-            image = image.convert("RGBA")
+            self.image = image.convert("RGBA")
+        else:
+            self.image = image
         self.create_window()
         self.penWidth = 3
         self.arrowSize = 10
@@ -44,7 +47,6 @@ class PinPicture(tk.Frame):
     # 创建窗口
     def create_window(self):
         self.master.overrideredirect(True)
-        self.master.attributes('-transparentcolor', 'white')
         self.master.config(bg='white')
         self.master.attributes('-topmost', True)
         self.canvas = tk.Canvas(self.master, height=self.image.height, width=self.image.width)
@@ -53,6 +55,8 @@ class PinPicture(tk.Frame):
         self.arrowBtn = tk.Button(text='箭头', command=self.set_arrow)
         self.lineBtn = tk.Button(text='直线', command=self.set_line)
         self.pinBtn = tk.Button(text='确定', command=self.pin_mode_event)
+        self.drawModeBtn = tk.Button(text="编辑", command=self.draw_mode_event)
+        self.detectBtn = tk.Button(text="识别", command=self.detect)
 
         self.master.bind("<Escape>", self.close)
         self.master.bind("<Enter>", self.enter_canvas)
@@ -69,7 +73,7 @@ class PinPicture(tk.Frame):
     def on_save(self, event):
         output = io.BytesIO()
         self.image.convert("RGB").save(output, "BMP")
-        data = output.getvalue()[14:]  # BMP 文件头前14字节
+        data = output.getvalue()[14:]  # BMP 文件头前14bit
         output.close()
         win32clipboard.OpenClipboard()
         try:
@@ -112,16 +116,15 @@ class PinPicture(tk.Frame):
     # 鼠标进入
     def enter_canvas(self, event):
         self.canvas.focus_set()
-        if self.drawModeBtn is None and not self.drawMode:
+        if not self.drawMode:
-            self.drawModeBtn = tk.Button(text="编辑", command=self.draw_mode_event)
-            self.drawModeBtn.pack(side=tk.TOP, expand=True)
-        elif not self.drawMode:
             self.drawModeBtn.pack(side=tk.TOP, expand=True)
+            self.detectBtn.pack(side=tk.TOP, expand=True)
 
     # 鼠标离开
     def leave(self, event):
         if self.drawModeBtn is not None and not self.drawMode:
             self.drawModeBtn.pack_forget()
+            self.detectBtn.pack_forget()
 
     # 进入编辑模式
     def draw_mode_event(self):
@@ -262,7 +265,7 @@ class PinPicture(tk.Frame):
         cosa = (x2 - x1) / l
         sina = (y2 - y1) / l
         d = self.arrowSize * (sina + cosa)
-        if (sina < cosa):
+        if sina < cosa:
             xa = x2 - d
             ya = y2 + math.sqrt(2 * self.arrowSize * self.arrowSize - d * d)
             xb = x2 - math.sqrt(2 * self.arrowSize * self.arrowSize - d * d)
@@ -291,3 +294,11 @@ class PinPicture(tk.Frame):
             self.image = Image.alpha_composite(self.image, self.pad)
             self.pad = Image.new('RGBA', self.image.size)
             self.drawPad = ImageDraw.Draw(self.pad)
+
+    def detect(self):
+        reader = easyocr.Reader(['ch_sim', 'en'])
+        result = reader.readtext(np.array(self.image), detail=0)
+        text = ''
+        for str in result:
+            text += str
+        print (text)