UnboundLocalError排查：Aadhaar掩码代码变量未定义问题

问题背景 #

这段用于Aadhaar卡号掩码的Python代码执行时触发UnboundLocalError，提示local variable 'final_image' referenced before assignment；尝试修改final_image后，又出现local variable 'x' referenced before assignment错误。

原始代码 #

def aadhar_mask_and_ocr(thres_image, resized_image):
    d = pytesseract.image_to_data(thres_image, output_type=pytesseract.Output.DICT)
    keys = list(d.keys())
    number_pattern = r"(?<!\d)\d{4}(?!\d)"
    n_boxes = len(d['text'])
    c = 0
    temp = []
    UID = []
    for i in range(n_boxes):
        if int(d['conf'][i]) > 20:
            if re.match(number_pattern, d['text'][i]):
                if c<2:
                    (x, y, w, h) = (d['left'][i], d['top'][i], d['width'][i], d['height'][i])
                    final_image = cv2.rectangle(resized_image, (x, y), (x + w, y + h), (255, 255, 255), -1)
                    temp.append(d['text'][i])
                    c+=1
                elif (c>=2) and (d['text'][i] in temp):
                    (x, y, w, h) = (d['left'][i], d['top'][i], d['width'][i], d['height'][i])
                    final_image = cv2.rectangle(resized_image, (x, y), (x + w, y + h), (255, 255, 255), -1)
                elif c==2:
                    UID = temp + [ d['text'][i]]
                    c+=1
                else:
                    continue
    final_image = cv2.resize(final_image, None, fx=0.33, fy=0.33)
    return final_image, UID

# Command-line argument
# ap = argparse.ArgumentParser()
# ap.add_argument("-i", "--image", required = True,
#     help = "Path to the image to be scanned")
# args = vars(ap.parse_args())
image = cv2.imread('/content/img1.jpeg')

thres_image, resized_image = preprocessing(image)
masked_image, UID = aadhar_mask_and_ocr(thres_image, resized_image)

报错信息 #

UnboundLocalError                         Traceback (most recent call last)
<ipython-input-44-231e4bfbf02b> in <module>
      7 
      8 thres_image, resized_image = preprocessing(image)
----> 9 masked_image, UID = aadhar_mask_and_ocr(thres_image, resized_image)
     10 
     11 print('Masked digits in given image. Displaying...')

<ipython-input-43-131f3ed0ec89> in aadhar_mask_and_ocr(thres_image, resized_image)
     23                 else:
     24                     continue
---> 25     final_image = cv2.resize(final_image, None, fx=0.33, fy=0.33)
     26     return final_image, UID

UnboundLocalError: local variable 'final_image' referenced before assignment

错误原因分析 #

1. final_image未定义错误：
   只有当循环中识别到符合条件的数字（置信度>20且匹配4位数字模式）时，final_image才会被赋值。如果没有任何符合条件的识别结果，循环结束后final_image从未被初始化，执行cv2.resize时就会触发未定义错误。

2. x未定义错误（修改后出现）：
   大概率是修改final_image时，误改了代码分支逻辑，导致某个需要使用x的分支没有先对x,y,w,h进行赋值，比如新增的分支直接调用cv2.rectangle但没先从d中提取坐标。

解决方案 #

1. 初始化final_image #

在函数开头就将final_image初始化为resized_image的副本，确保无论循环是否触发赋值，变量都存在：

final_image = resized_image.copy()  # 用copy避免修改原图像

2. 确保所有坐标变量的赋值逻辑覆盖 #

所有调用cv2.rectangle的分支必须先完成x,y,w,h的赋值，避免遗漏。

3. 优化条件判断逻辑 #

调整分支顺序，避免逻辑漏洞，比如将c==2的判断放在c>=2之前，防止逻辑被覆盖。

修改后的完整代码 #

import cv2
import pytesseract
import re

def aadhar_mask_and_ocr(thres_image, resized_image):
    # 初始化final_image为输入图像的副本，避免未赋值错误
    final_image = resized_image.copy()
    d = pytesseract.image_to_data(thres_image, output_type=pytesseract.Output.DICT)
    number_pattern = r"(?<!\d)\d{4}(?!\d)"
    n_boxes = len(d['text'])
    c = 0
    temp = []
    UID = []
    
    for i in range(n_boxes):
        # 跳过置信度低的识别结果
        if int(d['conf'][i]) <= 20:
            continue
        # 匹配4位数字段
        if not re.match(number_pattern, d['text'][i]):
            continue
            
        if c < 2:
            # 提取坐标并掩码
            x, y, w, h = d['left'][i], d['top'][i], d['width'][i], d['height'][i]
            cv2.rectangle(final_image, (x, y), (x + w, y + h), (255, 255, 255), -1)
            temp.append(d['text'][i])
            c += 1
        elif c == 2:
            # 收集UID第三段
            UID = temp + [d['text'][i]]
            c += 1
        elif c >= 2 and d['text'][i] in temp:
            # 掩码重复出现的前两段数字
            x, y, w, h = d['left'][i], d['top'][i], d['width'][i], d['height'][i]
            cv2.rectangle(final_image, (x, y), (x + w, y + h), (255, 255, 255), -1)
    
    final_image = cv2.resize(final_image, None, fx=0.33, fy=0.33)
    return final_image, UID

# 测试代码
image = cv2.imread('/content/img1.jpeg')
thres_image, resized_image = preprocessing(image)
masked_image, UID = aadhar_mask_and_ocr(thres_image, resized_image)

内容的提问来源于stack exchange，提问作者Vinay Kumar