Python正则提取OCR处理的银行PDF脱敏账号问题

Let's break down why your current regex isn't working and get you the correct account number extraction.

What's Wrong with the Original Regex? #

Your regex r'^[A-Z].*([0-9]{4}$)' is trying to match lines starting with an uppercase letter, then capture the last 4 digits of the line. But this doesn't target the specific structure of your account number: 8 X's followed by 4 digits (e.g., XXXXXXXX1002). It also only captures the final 4 digits, not the full 12-character account string you need.

Correct Regex Approach #

Since you know the account number is a fixed 12-character string with exactly 8 leading X's and 4 trailing digits, we can write a regex that directly matches this pattern:

import re

splits = [
    'ACCOUNT TYPE A/C. BALANCE (I) | FIXED DEPOSITS (LINKED) BAL. (II) TOTAL BALANCE (I+II) NOMINATION',
    'Savings A/c XXXXXXXX1002 4,58,636.08 0.00 4,58,636.08 Not Registered',
    'TOTAL 4,58,636.08 0.00 4,58,636.08'
]

account_number = None
for line in splits:
    # Match exactly 8 X's followed by 4 digits
    match = re.search(r'XXXXXXXX\d{4}', line)
    if match:
        account_number = match.group()
        break  # Stop searching once we find the account

print(account_number)  # Output: XXXXXXXX1002

Key Improvements: #

• The regex r'XXXXXXXX\d{4}' directly targets the unique structure of your desensitized account number, ensuring we only capture the exact string you need.
• Using re.search() looks for the pattern anywhere in the line (since the account number isn't at the start or end of the line in your example).
• We use match.group() to get the full matched string, not just a subset.

Flexible Alternative (If X Count Varies) #

If for some reason the number of leading X's isn't always 8 (but the total length is 12 with last 4 digits), you can use this more flexible regex:

match = re.search(r'X{8}\d{4}', line)
# Or, if X count could vary but total length is 12:
match = re.search(r'X+\d{4}', line)

This should reliably extract your account number from the OCR'd lines.

内容的提问来源于stack exchange，提问作者Subhojyoti Lahiri