PaddleOCR识别视频中的文字

前言

最近在爬取完视频数据后需要提取视频中的文本内容，搜寻一番没有找到满意的方案。最后自己总结出利用PaddleOCR提取视频文字的方案，简介易懂可实操。

1.PaddleOCR环境搭建

可参考其它环境搭建文档，这里推荐官方文档

个人心得：环境搭建过程坑较多，在anaconda虚拟环境下搭建更易成功。

2.提取视频帧，保存为图片集

根据视频情况调整截取视频帧的频率，这里是一秒一张图，保存图片到pics文件夹

传入视频所在文件夹的路径

def extract_frames_from_folder(folder_path):
    # 设置视频文件路径和输出文件夹路径
    video_path = os.path.join(folder_path, "video.mp4")#改为你的视频名
    output_folder = os.path.join(folder_path, "pics")

    # 创建输出文件夹
    os.makedirs(output_folder, exist_ok=True)

    # 打开视频文件
    cap = cv2.VideoCapture(video_path)

    # 视频帧率
    fps = cap.get(cv2.CAP_PROP_FPS)

    # 读取视频帧并保存为图片
    frame_count = 0
    second_count = 0
    while True:
        ret, frame = cap.read()
        if not ret:
            break

        # 如果当前帧的时间大于等于当前秒数
        if frame_count >= fps * second_count:
            # 保存图片到输出文件夹
            frame_name = f"{second_count}.jpg"
            output_path = os.path.join(output_folder, frame_name)
            cv2.imwrite(output_path, frame)

            # 打印已保存的帧数
            print(f"Saved frame {frame_count} at second {second_count}")
            second_count += 1

        # 递增帧计数器
        frame_count += 1

    # 释放视频流
    cap.release()

3.识别图片文字

依次识别每张图片的文字，写入text_ocr.txt

def recognize_and_save_text(folder_path):
    pics_path = os.path.join(folder_path, "pics")
    text_file_path = os.path.join(folder_path, "text_ocr.txt")
    with open(text_file_path, 'a', encoding='utf-8') as f:
        pass
    # 初始化PaddleOCR
    ocr = PaddleOCR()

    # 用于存储已识别的文字内容
    recognized_text = set()

    # 获取文件夹中的所有图片文件，并按照数字顺序排序
    image_files = sorted(
        [filename for filename in os.listdir(pics_path) if filename.endswith('.jpg') or filename.endswith('.png')],
        key=lambda x: int(os.path.splitext(x)[0]))

    for filename in image_files:
        # 图片路径
        image_path = os.path.join(pics_path, filename)

        try:
            # 使用PaddleOCR识别文字
            result = ocr.ocr(image_path, cls=True)

            # 检查识别结果是否为空
            if result:
                # 提取识别结果中的文字内容
                text = ''.join([word[1][0] for line in result for word in line])

                # 检查识别的文字是否已经在集合中，若不在则加入集合并写入文本文件
                if text not in recognized_text:
                    recognized_text.add(text)
                    with open(text_file_path, 'a', encoding='utf-8') as f:
                        f.write(text + '\n')
            else:
                print(f"No text detected in {image_path}")

        except Exception as e:
            print(f"Error processing {image_path}: {e}")
            continue

    print("文字识别并保存完成。")

4.完整代码

import shutil
import cv2
import os
from paddleocr import PaddleOCR

def extract_frames_from_folder(folder_path):
    # 设置视频文件路径和输出文件夹路径
    video_path = os.path.join(folder_path, "video.mp4")
    output_folder = os.path.join(folder_path, "pics")

    # 创建输出文件夹
    os.makedirs(output_folder, exist_ok=True)

    # 打开视频文件
    cap = cv2.VideoCapture(video_path)

    # 视频帧率
    fps = cap.get(cv2.CAP_PROP_FPS)

    # 读取视频帧并保存为图片
    frame_count = 0
    second_count = 0
    while True:
        ret, frame = cap.read()
        if not ret:
            break

        # 如果当前帧的时间大于等于当前秒数
        if frame_count >= fps * second_count:
            # 保存图片到输出文件夹
            frame_name = f"{second_count}.jpg"
            output_path = os.path.join(output_folder, frame_name)
            cv2.imwrite(output_path, frame)

            # 打印已保存的帧数
            print(f"Saved frame {frame_count} at second {second_count}")
            second_count += 1

        # 递增帧计数器
        frame_count += 1

    # 释放视频流
    cap.release()


def recognize_and_save_text(folder_path):
    pics_path = os.path.join(folder_path, "pics")
    text_file_path = os.path.join(folder_path, "text_ocr.txt")
    with open(text_file_path, 'a', encoding='utf-8') as f:
        pass
    # 初始化PaddleOCR
    ocr = PaddleOCR()

    # 用于存储已识别的文字内容
    recognized_text = set()

    # 获取文件夹中的所有图片文件，并按照数字顺序排序
    image_files = sorted(
        [filename for filename in os.listdir(pics_path) if filename.endswith('.jpg') or filename.endswith('.png')],
        key=lambda x: int(os.path.splitext(x)[0]))

    for filename in image_files:
        # 图片路径
        image_path = os.path.join(pics_path, filename)

        try:
            # 使用PaddleOCR识别文字
            result = ocr.ocr(image_path, cls=True)

            # 检查识别结果是否为空
            if result:
                # 提取识别结果中的文字内容
                text = ''.join([word[1][0] for line in result for word in line])

                # 检查识别的文字是否已经在集合中，若不在则加入集合并写入文本文件
                if text not in recognized_text:
                    recognized_text.add(text)
                    with open(text_file_path, 'a', encoding='utf-8') as f:
                        f.write(text + '\n')
            else:
                print(f"No text detected in {image_path}")

        except Exception as e:
            print(f"Error processing {image_path}: {e}")
            continue

    print("文字识别并保存完成。")


def check_pics_folder(folder_path):
    """
    检查指定文件夹中是否存在名为'pics'的子文件夹。

    参数:
    folder_path (str): 要检查的文件夹路径。

    返回:
    bool: 如果存在名为'pics'的子文件夹则返回True，否则返回False。
    """
    # 判断文件夹是否存在
    if os.path.exists(folder_path):
        # 获取文件夹下所有子文件夹的列表
        subfolders = next(os.walk(folder_path))[1]

        # 判断是否存在名为"pics"的文件夹
        if "pics" in subfolders:
            return True
        else:
            return False
    else:
        print("指定的文件夹路径不存在。")
        return False


main_folder = 'data'
extract_frames_from_folder(main_folder)
recognize_and_save_text(main_folder)
#提取完后删除图片文件夹
#shutil.rmtree(os.path.join(main_folder, 'pics'))