预览
第1步:理解基本结构和导入必要的库
python
# 1. 首先导入需要的库
import os # 用于处理文件和路径
import cv2 # 用于图像处理
import numpy as np # 用于数值计算
from paddleocr import PaddleOCR # 用于文字识别
from pdf2image import convert_from_path # 用于PDF转图像
import time # 用于计时
第2步:创建基本类结构
python
class PDFTextExtractor:
def __init__(self):
# 初始化OCR工具
self.ocr = PaddleOCR(
use_angle_cls=True,
lang='ch', # 中文识别
use_gpu=False, # 不使用GPU
show_log=False # 不显示日志
)
# 定义要识别的颜色范围(黄色和红色)
self.color_ranges = {
'yellow': {
'lower': np.array([15, 70, 70]),
'upper': np.array([35, 255, 255])
},
'red': {
'lower': np.array([0, 70, 70]),
'upper': np.array([15, 255, 255])
}
}
第3步:创建主要处理函数
python
def process_pdf(self, pdf_path, output_path='extracted_text.txt'):
try:
# 检查PDF文件是否存在
if not os.path.exists(pdf_path):
raise FileNotFoundError(f"PDF文件不存在: {pdf_path}")
print(f"开始处理PDF: {pdf_path}")
start_time = time.time()
# 设置poppler路径(需要先安装poppler)
poppler_path = r"E:\Proper\poppler-24.08.0\Library\bin"
if not os.path.exists(poppler_path):
raise Exception(f"Poppler 路径不存在: {poppler_path}")
# 获取PDF总页数
total_pages = self.get_pdf_page_count(pdf_path, poppler_path)
print(f"PDF总页数: {total_pages}")
# 处理每一页
with open(output_path, 'w', encoding='utf-8') as f:
# 处理代码...
第4步:创建图像预处理函数
python
def preprocess_image(self, pil_image):
"""图像预处理函数"""
# 1. 调整图像大小
pil_image = self.resize_image(pil_image)
# 2. 转换为OpenCV格式并预处理
img = cv2.cvtColor(np.array(pil_image), cv2.COLOR_RGB2BGR) # 转换颜色空间
img = cv2.GaussianBlur(img, (3, 3), 0) # 使用高斯模糊降噪
img = cv2.convertScaleAbs(img, alpha=1.2, beta=10) # 调整对比度和亮度
return img
def resize_image(self, image):
"""调整图像大小的函数"""
width, height = image.size
max_dimension = 2000 # 设置最大尺寸
# 如果图像太大,就等比例缩小
if width > max_dimension or height > max_dimension:
scale = max_dimension / max(width, height)
new_width = int(width * scale)
new_height = int(height * scale)
return image.resize((new_width, new_height))
return image
第5步:创建文本提取函数
python
def extract_colored_text(self, img, color_lower, color_upper):
"""提取特定颜色区域的文本"""
try:
# 1. 转换为HSV颜色空间(更容易处理颜色)
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# 2. 创建颜色掩码
mask = cv2.inRange(hsv, color_lower, color_upper)
# 3. 图像处理优化
kernel = np.ones((3, 3), np.uint8)
mask = cv2.dilate(mask, kernel, iterations=2) # 膨胀
mask = cv2.erode(mask, kernel, iterations=1) # 腐蚀
mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel) # 开运算
# 4. 提取颜色区域
result = cv2.bitwise_and(img, img, mask=mask)
# 5. 转换为灰度图
gray = cv2.cvtColor(result, cv2.COLOR_BGR2GRAY)
gray = cv2.convertScaleAbs(gray, alpha=1.5, beta=10)
# 6. OCR文字识别
ocr_result = self.ocr.ocr(gray, cls=True)
# 7. 处理识别结果
texts = []
if ocr_result is not None:
for line in ocr_result:
if line is not None:
for word_info in line:
if isinstance(word_info, list) and len(word_info) >= 2:
text = word_info[1][0].strip()
confidence = word_info[1][1]
if confidence > 0.6 and text: # 置信度大于0.6的结果
texts.append(text)
return texts
except Exception as e:
print(f"文本提取错误: {str(e)}")
return []
第6步:创建PDF页数获取函数
python
def get_pdf_page_count(self, pdf_path, poppler_path):
"""获取PDF总页数"""
try:
# 使用pdf2image的方法获取PDF信息
from pdf2image.pdf2image import pdfinfo_from_path
info = pdfinfo_from_path(pdf_path, poppler_path=poppler_path)
return info['Pages']
except Exception as e:
print(f"获取PDF页数失败: {str(e)}")
return 0
第7步:创建主函数
python
def main():
try:
# 1. 创建PDF文本提取器实例
extractor = PDFTextExtractor()
# 2. 设置PDF文件路径
pdf_path = r"你的PDF文件路径"
# 3. 处理PDF
extractor.process_pdf(pdf_path)
except Exception as e:
print(f"程序执行失败: {str(e)}")
# 程序入口
if __name__ == "__main__":
main()
使用步骤:
- 安装必要的库:
bash
pip install opencv-python numpy paddleocr pdf2image pillow
-
安装Poppler:
- 下载Poppler并解压
- 将bin目录添加到环境变量
-
修改代码中的路径:
- PDF文件路径
- Poppler路径
-
运行程序
好的,让我详细解释如何一步步实现这个程序:
第8步:环境准备
- 创建项目文件夹
bash
mkdir pdf_extractor
cd pdf_extractor
- 创建虚拟环境(推荐)
bash
python -m venv venv
# Windows激活虚拟环境
venv\Scripts\activate
# Linux/Mac激活虚拟环境
source venv/bin/activate
- 安装所需库
bash
pip install opencv-python
pip install numpy
pip install paddleocr
pip install pdf2image
pip install pillow
第9步:创建项目结构
pdf_extractor/
├── main.py # 主程序文件
├── requirements.txt # 依赖库列表
└── output/ # 输出文件夹
第10步:编写完整代码
在 main.py
中:
python
import os
import cv2
import numpy as np
from paddleocr import PaddleOCR
from pdf2image import convert_from_path
import time
class PDFTextExtractor:
def __init__(self):
# 初始化OCR
print("初始化OCR系统...")
self.ocr = PaddleOCR(
use_angle_cls=True,
lang='ch',
use_gpu=False,
show_log=False
)
# 定义颜色范围
self.color_ranges = {
'yellow': {
'lower': np.array([15, 70, 70]),
'upper': np.array([35, 255, 255])
},
'red': {
'lower': np.array([0, 70, 70]),
'upper': np.array([15, 255, 255])
}
}
print("初始化完成!")
def process_pdf(self, pdf_path, output_path='output/extracted_text.txt'):
"""处理PDF文件的主函数"""
try:
# 创建输出目录
os.makedirs(os.path.dirname(output_path), exist_ok=True)
# 检查文件是否存在
if not os.path.exists(pdf_path):
raise FileNotFoundError(f"PDF文件不存在: {pdf_path}")
print(f"开始处理PDF: {pdf_path}")
start_time = time.time()
# 处理每一页
self._process_pages(pdf_path, output_path)
# 完成处理
end_time = time.time()
print(f"\n处理完成!用时: {end_time - start_time:.2f}秒")
print(f"结果已保存到: {output_path}")
except Exception as e:
print(f"处理过程中出错: {str(e)}")
raise
def _process_pages(self, pdf_path, output_path):
"""处理PDF的所有页面"""
# 获取PDF总页数
total_pages = self.get_pdf_page_count(pdf_path)
print(f"PDF总页数: {total_pages}")
# 打开输出文件
with open(output_path, 'w', encoding='utf-8') as f:
# 处理每一页
for page_num in range(1, total_pages + 1):
self._process_single_page(pdf_path, page_num, total_pages, f)
好的,让我们继续完成代码的其余部分:
python
def _process_single_page(self, pdf_path, page_num, total_pages, output_file):
"""处理单个PDF页面"""
print(f"\n处理第 {page_num}/{total_pages} 页...")
try:
# 1. 转换PDF页面为图像
pages = convert_from_path(
pdf_path,
first_page=page_num,
last_page=page_num,
dpi=200, # 设置分辨率
poppler_path=r"E:\Proper\poppler-24.08.0\Library\bin", # 修改为你的poppler路径
thread_count=1
)
if not pages:
print(f"警告: 第 {page_num} 页转换失败")
return
# 2. 获取页面图像
page = pages[0]
# 3. 预处理图像
img = self.preprocess_image(page)
# 4. 处理每种颜色
page_results = []
for color_name, color_range in self.color_ranges.items():
print(f"处理{color_name}色文本...")
highlighted_text = self.extract_colored_text(
img.copy(),
color_range['lower'],
color_range['upper']
)
if highlighted_text:
page_results.extend(highlighted_text)
# 5. 保存结果
if page_results:
output_file.write(f"\n第{page_num}页标注文本:\n")
output_file.write('\n'.join(page_results) + '\n')
output_file.flush()
print(f"第 {page_num} 页找到 {len(page_results)} 条文本")
else:
print(f"第 {page_num} 页未找到高亮文本")
# 6. 清理内存
del pages
del page
del img
except Exception as e:
print(f"处理第 {page_num} 页时出错: {str(e)}")
def preprocess_image(self, pil_image):
"""图像预处理"""
try:
# 1. 调整图像大小
resized_image = self.resize_image(pil_image)
# 2. 转换为OpenCV格式
img = cv2.cvtColor(np.array(resized_image), cv2.COLOR_RGB2BGR)
# 3. 图像增强
img = cv2.GaussianBlur(img, (3, 3), 0) # 降噪
img = cv2.convertScaleAbs(img, alpha=1.2, beta=10) # 增加对比度和亮度
return img
except Exception as e:
print(f"图像预处理错误: {str(e)}")
raise
def resize_image(self, image):
"""调整图像大小"""
try:
width, height = image.size
max_dimension = 2000
# 如果图像太大,进行缩放
if width > max_dimension or height > max_dimension:
scale = max_dimension / max(width, height)
new_width = int(width * scale)
new_height = int(height * scale)
return image.resize((new_width, new_height))
return image
except Exception as e:
print(f"图像缩放错误: {str(e)}")
raise
使用示例:
python
def main():
try:
# 1. 创建输出目录
os.makedirs('output', exist_ok=True)
# 2. 创建提取器实例
print("初始化PDF文本提取器...")
extractor = PDFTextExtractor()
# 3. 设置PDF文件路径
pdf_path = r"你的PDF文件路径" # 修改为你的PDF文件路径
output_path = "output/extracted_text.txt"
# 4. 处理PDF
print(f"开始处理PDF文件: {pdf_path}")
extractor.process_pdf(pdf_path, output_path)
except Exception as e:
print(f"程序执行失败: {str(e)}")
raise
if __name__ == "__main__":
main()
使用说明:
-
准备工作:
- 安装所需库
- 安装Poppler并设置路径
- 准备要处理的PDF文件
-
修改配置:
- 修改PDF文件路径
- 修改Poppler路径
- 根据需要调整颜色范围
-
运行程序:
bash
python main.py
- 查看结果 :
- 输出文件将保存在output目录下
- 程序会显示处理进度和结果
完整项目代码
python
import os
import cv2
import numpy as np
from paddleocr import PaddleOCR
from pdf2image import convert_from_path
import time
class PDFTextExtractor:
def __init__(self):
self.ocr = PaddleOCR(
use_angle_cls=True,
lang='ch',
use_gpu=False,
show_log=False
)
self.color_ranges = {
'yellow': {
'lower': np.array([15, 70, 70]),
'upper': np.array([35, 255, 255])
},
'red': {
'lower': np.array([0, 70, 70]),
'upper': np.array([15, 255, 255])
}
}
def process_pdf(self, pdf_path, output_path='extracted_text.txt'):
try:
if not os.path.exists(pdf_path):
raise FileNotFoundError(f"PDF文件不存在: {pdf_path}")
print(f"开始处理PDF: {pdf_path}")
start_time = time.time()
poppler_path = r"E:\Proper\poppler-24.08.0\Library\bin"
if not os.path.exists(poppler_path):
raise Exception(f"Poppler 路径不存在: {poppler_path}")
# 获取PDF总页数
total_pages = self.get_pdf_page_count(pdf_path, poppler_path)
print(f"PDF总页数: {total_pages}")
with open(output_path, 'w', encoding='utf-8') as f:
for page_num in range(1, total_pages + 1):
print(f"\n处理第 {page_num}/{total_pages} 页...")
try:
pages = convert_from_path(
pdf_path,
first_page=page_num,
last_page=page_num,
dpi=200,
poppler_path=poppler_path,
thread_count=1
)
if not pages:
print(f"警告: 第 {page_num} 页转换失败")
continue
page = pages[0]
# 转换和预处理图像
img = self.preprocess_image(page)
# 处理每种颜色
page_results = []
for color_name, color_range in self.color_ranges.items():
print(f"处理{color_name}色文本...")
highlighted_text = self.extract_colored_text(
img.copy(), # 使用图像副本
color_range['lower'],
color_range['upper']
)
if highlighted_text:
page_results.extend(highlighted_text)
# 保存结果
if page_results:
f.write(f"\n第{page_num}页标注文本:\n")
f.write('\n'.join(page_results) + '\n')
f.flush()
print(f"第 {page_num} 页找到 {len(page_results)} 条文本")
else:
print(f"第 {page_num} 页未找到高亮文本")
# 清理内存
del pages
del page
del img
except Exception as e:
print(f"处理第 {page_num} 页时出错: {str(e)}")
continue
end_time = time.time()
print(f"\n处理完成!用时: {end_time - start_time:.2f}秒")
print(f"结果已保存到: {output_path}")
except Exception as e:
print(f"处理过程中出错: {str(e)}")
raise
def preprocess_image(self, pil_image):
"""图像预处理"""
# 调整大小
pil_image = self.resize_image(pil_image)
# 转换为OpenCV格式并预处理
img = cv2.cvtColor(np.array(pil_image), cv2.COLOR_RGB2BGR)
img = cv2.GaussianBlur(img, (3, 3), 0) # 降噪
img = cv2.convertScaleAbs(img, alpha=1.2, beta=10) # 增加对比度和亮度
return img
def get_pdf_page_count(self, pdf_path, poppler_path):
"""获取PDF页数"""
try:
pages = convert_from_path(
pdf_path,
dpi=72,
poppler_path=poppler_path,
first_page=1,
last_page=1
)
# 使用 pdf2image 的方法获取总页数
from pdf2image.pdf2image import pdfinfo_from_path
info = pdfinfo_from_path(pdf_path, poppler_path=poppler_path)
return info['Pages']
except Exception as e:
print(f"获取PDF页数失败: {str(e)}")
return 0
def resize_image(self, image):
"""调整图像大小"""
width, height = image.size
max_dimension = 2000 # 增加最大尺寸以提高识别率
if width > max_dimension or height > max_dimension:
scale = max_dimension / max(width, height)
new_width = int(width * scale)
new_height = int(height * scale)
return image.resize((new_width, new_height))
return image
def extract_colored_text(self, img, color_lower, color_upper):
"""提取特定颜色区域的文本"""
try:
# 转换颜色空间
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# 创建掩码
mask = cv2.inRange(hsv, color_lower, color_upper)
# 形态学操作
kernel = np.ones((3, 3), np.uint8)
mask = cv2.dilate(mask, kernel, iterations=2)
mask = cv2.erode(mask, kernel, iterations=1)
mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
# 应用掩码
result = cv2.bitwise_and(img, img, mask=mask)
# 转换为灰度图
gray = cv2.cvtColor(result, cv2.COLOR_BGR2GRAY)
# 增强对比度
gray = cv2.convertScaleAbs(gray, alpha=1.5, beta=10)
# 保存处理后的图像用于调试
# cv2.imwrite(f'debug_page_{time.time()}.png', gray)
# OCR识别
ocr_result = self.ocr.ocr(gray, cls=True)
# 处理OCR结果
texts = []
if ocr_result is not None: # 添加空值检查
for line in ocr_result:
if line is not None: # 添加行级空值检查
for word_info in line:
if isinstance(word_info, list) and len(word_info) >= 2:
text = word_info[1][0].strip()
confidence = word_info[1][1]
if confidence > 0.6 and text:
texts.append(text)
return texts
except Exception as e:
print(f"文本提取错误: {str(e)}")
return []
def main():
try:
extractor = PDFTextExtractor()
pdf_path = r"E:\z_library_books\平时作业\中国旅游文化_11608595(1).pdf"
extractor.process_pdf(pdf_path)
except Exception as e:
print(f"程序执行失败: {str(e)}")
if __name__ == "__main__":
main()
补充说明:便于理解程序的运行流程
好的,让我更详细地解释每个步骤:
1. 程序启动和初始化
python
def main():
try:
# 1. 创建PDF文本提取器
extractor = PDFTextExtractor()
# 这一步会:
# - 启动PaddleOCR引擎(文字识别工具)
# - 设置识别中文
# - 设置不使用GPU
# - 设置要识别的颜色范围(黄色和红色)
2. 设置文件路径和开始处理
python
# 2. 设置PDF文件路径
pdf_path = r"E:\z_library_books\平时作业\中国旅游文化_11608595(1).pdf"
# 3. 开始处理PDF
extractor.process_pdf(pdf_path)
3. PDF处理流程(process_pdf函数)
python
def process_pdf(self, pdf_path, output_path='extracted_text.txt'):
try:
# 1. 检查PDF文件是否存在
if not os.path.exists(pdf_path):
raise FileNotFoundError("PDF文件不存在")
# 2. 记录开始时间
start_time = time.time()
# 3. 设置poppler工具路径(用于转换PDF为图片)
poppler_path = r"E:\Proper\poppler-24.08.0\Library\bin"
# 4. 获取PDF总页数
total_pages = self.get_pdf_page_count(pdf_path, poppler_path)
print(f"PDF总页数: {total_pages}")
# 5. 创建输出文件
with open(output_path, 'w', encoding='utf-8') as f:
# 6. 逐页处理
for page_num in range(1, total_pages + 1):
# 处理每一页...
4. 单页处理流程
python
# 对于每一页:
try:
# 1. 将PDF页面转换为图片
pages = convert_from_path(
pdf_path,
first_page=page_num,
last_page=page_num,
dpi=200, # 设置图片清晰度
poppler_path=poppler_path
)
# 2. 获取页面图片
page = pages[0]
# 3. 预处理图片
img = self.preprocess_image(page)
# - 调整图片大小
# - 增加清晰度
# - 调整亮度和对比度
# 4. 处理每种颜色
page_results = []
for color_name, color_range in self.color_ranges.items():
print(f"处理{color_name}色文本...")
# 提取特定颜色的文本
highlighted_text = self.extract_colored_text(
img.copy(),
color_range['lower'],
color_range['upper']
)
if highlighted_text:
page_results.extend(highlighted_text)
# 5. 保存这一页的结果
if page_results:
f.write(f"\n第{page_num}页标注文本:\n")
f.write('\n'.join(page_results) + '\n')
5. 文本提取流程(extract_colored_text函数)
python
def extract_colored_text(self, img, color_lower, color_upper):
try:
# 1. 转换颜色空间,便于找到高亮部分
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# 2. 创建掩码(找出高亮部分)
mask = cv2.inRange(hsv, color_lower, color_upper)
# 3. 优化掩码
kernel = np.ones((3, 3), np.uint8)
mask = cv2.dilate(mask, kernel, iterations=2)
mask = cv2.erode(mask, kernel, iterations=1)
# 4. 提取高亮区域
result = cv2.bitwise_and(img, img, mask=mask)
# 5. 转为灰度图
gray = cv2.cvtColor(result, cv2.COLOR_BGR2GRAY)
# 6. OCR识别文字
ocr_result = self.ocr.ocr(gray, cls=True)
# 7. 处理识别结果
texts = []
if ocr_result is not None:
for line in ocr_result:
if line is not None:
for word_info in line:
text = word_info[1][0].strip()
confidence = word_info[1][1]
if confidence > 0.6 and text:
texts.append(text)
return texts
这个程序就像一个阅读助手:
- 先准备好工具(OCR引擎)
- 打开PDF文件
- 一页一页地:
- 把PDF页面转成图片
- 找出高亮的部分
- 识别高亮部分的文字
- 记录下识别到的文字
- 最后把所有记录的文字保存到文件中