report-detect/archive/temp_scripts/analyze_ydq_v2.py

"""
Debug CMA extraction - handle new PaddleOCR format.
"""
import os
import cv2
import numpy as np
import re

# Set environment variables
os.environ['PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK'] = 'True'

from paddleocr import PaddleOCR

# Initialize OCR
print('Initializing PaddleOCR...')
ocr = PaddleOCR(use_angle_cls=True, lang='ch')

# Read image
img = cv2.imread('debug_images/YDQ25_002294_page1.png')
h, w = img.shape[:2]
print(f'Image size: {w}x{h}')

# Extract top-right area
top_right = img[0:int(h*0.4), int(w*0.4):w]
print(f'Top-right area: {top_right.shape[1]}x{top_right.shape[0]}')

# OCR on top-right
print('\nRunning OCR on top-right area...')
result = ocr.ocr(top_right)

print(f'OCR result type: {type(result)}')

# Handle new PaddleOCR format (dict with rec_texts)
rec_texts = []
rec_scores = []

if isinstance(result, dict):
    print('OCR returned dict format (new API)')
    rec_texts = result.get('rec_texts', [])
    rec_scores = result.get('rec_scores', [])
    print(f'Found {len(rec_texts)} text lines')
    for i, text in enumerate(rec_texts):
        print(f'{i+1}. {text}')
elif isinstance(result, list) and len(result) > 0:
    print('OCR returned list format (old API)')
    if isinstance(result[0], dict):
        rec_texts = result[0].get('rec_texts', [])
        rec_scores = result[0].get('rec_scores', [])
    elif isinstance(result[0], list):
        for line in result[0]:
            if len(line) >= 2:
                text = line[1][0] if isinstance(line[1], (list, tuple)) else str(line[1])
                rec_texts.append(text)

# Find 11-12 digit numbers
cma_pattern = re.compile(r'\d{11,12}')
all_numbers = []

for i, text in enumerate(rec_texts):
    cleaned = text.replace(' ', '').replace('-', '').replace(':', '')
    matches = cma_pattern.findall(cleaned)
    for match in matches:
        all_numbers.append({
            'number': match,
            'text': text
        })

print(f'\nFound {len(all_numbers)} 11-digit numbers in top-right:')
for i, num_info in enumerate(all_numbers, 1):
    print(f'{i}. {num_info["number"]} - Text: "{num_info["text"]}"')

expected = '240020349096'
found = any(n['number'] == expected for n in all_numbers)
print(f'\nExpected CMA {expected}: {"FOUND" if found else "NOT FOUND"}')

# Full page OCR
print('\n' + '='*80)
print('Running full page OCR...')
full_result = ocr.ocr(img)

full_rec_texts = []
if isinstance(full_result, dict):
    full_rec_texts = full_result.get('rec_texts', [])
elif isinstance(full_result, list) and len(full_result) > 0:
    if isinstance(full_result[0], dict):
        full_rec_texts = full_result[0].get('rec_texts', [])
    elif isinstance(full_result[0], list):
        for line in full_result[0]:
            if len(line) >= 2:
                text = line[1][0] if isinstance(line[1], (list, tuple)) else str(line[1])
                full_rec_texts.append(text)

print(f'Found {len(full_rec_texts)} text lines on full page')

# Find all 11-digit numbers
all_numbers_full = []
for text in full_rec_texts:
    cleaned = text.replace(' ', '').replace('-', '').replace(':', '')
    matches = cma_pattern.findall(cleaned)
    for match in matches:
        all_numbers_full.append({
            'number': match,
            'text': text
        })

print(f'\nFound {len(all_numbers_full)} 11-digit numbers on full page:')
print('First 20:')
for i, num_info in enumerate(all_numbers_full[:20], 1):
    text_preview = num_info["text"][:80]
    print(f'{i}. {num_info["number"]} - Text: "{text_preview}"')

found_full = any(n['number'] == expected for n in all_numbers_full)
print(f'\nExpected CMA {expected} on full page: {"FOUND" if found_full else "NOT FOUND"}')

# Conclusion
print('\n' + '='*80)
print('ANALYSIS COMPLETE')
print('='*80)
if found_full:
    print(f'SUCCESS: Expected CMA {expected} was found')
else:
    print(f'FAILURE: Expected CMA {expected} was NOT found')
    print('\nPossible reasons:')
    print('1. CMA code is on a different page (not page 1)')
    print('2. CMA code is in a graphic/image that OCR cannot read')
    print('3. The CMA code format is different (not 11 digits)')
    print('4. The expected CMA code in results.json is incorrect')
    print('\nRecommendation: Check other pages of the PDF or verify the expected CMA code')