Объединение боксов происходит по блокам и линиям, а не только линиям

main.py

@@ -1,12 +1,16 @@
 import time
+import random
+import math
 import logging
 import numpy as np
+import pandas as pd
 import pygetwindow as gw
 import pyautogui
 from pytesseract import pytesseract
 from pytesseract import Output
 import cv2
 
+
 win_name = "EVE - Nostrom Stone"
 pytesseract.tesseract_cmd = "C:\\Program Files\\Tesseract-OCR\\tesseract.exe"
 
@@ -22,23 +26,32 @@ save_result = True
 
 def save_highlighted_screenshot(screenshot, boxes, filename):
     new_image = screenshot.copy()
-    for r in boxes.itertuples():
-        (x, y, w, h) = (
-            r.left,
-            r.top,
-            r.width,
-            r.height,
-        )
-        cv2.rectangle(new_image, (x, y), (x + w, y + h), (0, 255, 0), 2)
-        cv2.putText(
-            new_image,
-            f"'{r.text}': {r.left}.{r.top} {r.width}.{r.height}",
-            (x, y - 10),
-            cv2.FONT_HERSHEY_SIMPLEX,
-            0.4,
-            (0, 255, 0),
-            1,
+
+    for block in boxes["block_num"].unique():
+        boxes_in_block = boxes.loc[boxes["block_num"] == block]
+        # Повторяемая генерация rgb-цвета для числа
+        col = (
+            round(math.sin(0.024 * block * 255 / 3 + 0) * 127 + 128),
+            round(math.sin(0.024 * block * 255 / 3 + 2) * 127 + 128),
+            round(math.sin(0.024 * block * 255 / 3 + 4) * 127 + 128),
         )
+        for r in boxes_in_block.itertuples():
+            (x, y, w, h) = (
+                r.left,
+                r.top,
+                r.width,
+                r.height,
+            )
+            cv2.rectangle(new_image, (x, y), (x + w, y + h), col, 2)
+            cv2.putText(
+                new_image,
+                f"'{r.text}' x:{r.left} y:{r.top} l:{r.line_num}",
+                (x, y - 10),
+                cv2.FONT_HERSHEY_SIMPLEX,
+                0.4,
+                col,
+                1,
+            )
     cv2.imwrite(f"images/{filename}.png", new_image)
 
 
@@ -81,17 +94,19 @@ def get_boxes(screenshot):
         screenshot,
         lang="eng",
         output_type=Output.DATAFRAME,
-        config="--psm 4 -c preserve_interword_spaces=1",
+        config="--psm 3 -c preserve_interword_spaces=1",
     )
-    base_boxes = raw_boxes.loc[(raw_boxes["text"].str.len() > 3)][
-        ["left", "top", "width", "height", "text", "line_num", "word_num"]
-    ]
+
+    if save_result:
+        raw_boxes.to_excel("xlsx/0_raw_boxes.xlsx", index=False)
+
+    base_boxes = raw_boxes.loc[raw_boxes["text"].str.len() > 3]
 
     if save_result:
         save_highlighted_screenshot(
             screenshot, base_boxes, "1_base_highlighted_screenshot"
         )
-        base_boxes.to_excel("xlsx/base_boxes.xlsx", index=False)
+        base_boxes.to_excel("xlsx/1_base_boxes.xlsx", index=False)
 
     logging.info("Боксы получены")
 
@@ -99,29 +114,59 @@ def get_boxes(screenshot):
 
 
 def union_boxes(base_boxes):
-    grouped_words = base_boxes.groupby("line_num", as_index=False)
-    result_boxes = grouped_words["width"].sum()
-    result_boxes = result_boxes.merge(
-        grouped_words["height"].max(), on="line_num", how="left"
-    )
-    result_boxes = result_boxes.merge(
-        grouped_words["left"].min(), on="line_num", how="left"
-    )
-    result_boxes = result_boxes.merge(
-        grouped_words["top"].min(), on="line_num", how="left"
-    )
-    result_boxes = result_boxes.merge(
-        grouped_words["text"].apply(" ".join),
-        on="line_num",
-        how="left",
+    result_phrases = pd.DataFrame(
+        columns=[
+            "left",
+            "top",
+            "width",
+            "height",
+            "text",
+            "block_num",
+            "line_num",
+        ]
     )
+    for box in base_boxes["block_num"].unique():
+        words_in_blocks = base_boxes.loc[base_boxes["block_num"] == box]
+
+        grouped_words = words_in_blocks.groupby("line_num", as_index=False)
+
+        box_phrases = grouped_words["width"].sum()
+        box_phrases = box_phrases.merge(
+            grouped_words["height"].max(), on="line_num", how="left"
+        )
+        box_phrases = box_phrases.merge(
+            grouped_words["left"].min(), on="line_num", how="left"
+        )
+        box_phrases = box_phrases.merge(
+            grouped_words["top"].min(), on="line_num", how="left"
+        )
+        box_phrases = box_phrases.merge(
+            grouped_words["text"].apply(" ".join),
+            on="line_num",
+            how="left",
+        )
+        box_phrases["block_num"] = box
+
+        rightest_box = words_in_blocks.loc[
+            words_in_blocks["left"] == words_in_blocks["left"].max()
+        ]
+        leftest_box = words_in_blocks.loc[
+            words_in_blocks["left"] == words_in_blocks["left"].min()
+        ]
+        box_phrases["width"] = (
+            rightest_box.iloc[0].left
+            + rightest_box.iloc[0].width
+            - leftest_box.iloc[0].left
+        )
+
+        result_phrases = pd.concat([result_phrases, box_phrases])
 
     if save_result:
-        result_boxes.to_excel("xlsx/unioned_boxes.xlsx", index=False)
+        result_phrases.to_excel("xlsx/2_unioned_boxes.xlsx", index=False)
 
     logging.info("Боксы объединены")
 
-    return result_boxes
+    return result_phrases
 
 
 screenshot = get_screenshot()