python读取视频字幕功能，结合百度ai图片文字识别，opencv-python

所安装的python库

pip install opencv-python
pip install Pillow
pip install numpy

图片#灰度界限，大于这个值为黑色，小于这个值为白色
可以使用ps软件使用颜色拾取工具，其中RGB模式下的G值则为灰度值。

全部代码如下：

from cv2 import cv2 as cv
from PIL import Image
import base64,re
from io import BytesIO
from aip import AipOcr
import time

#""" 你的 APPID AK SK """
APP_ID = '百度id'
API_KEY = '百度key'
SECRET_KEY = '百度secret'
client = AipOcr(APP_ID, API_KEY, SECRET_KEY)



vedio_contents = [[]]

#""" 读取图片 """
def get_file_content(filePath):
    with open(filePath, 'rb') as fp:
        return fp.read()

def words_baidu_read(imagefile):
    try:
        image = get_file_content(imagefile)
        #""" 调用通用文字识别, 图片参数为本地图片 """
        jieguo = client.basicGeneral(image)
        words = jieguo['words_result'][0]['words']
        return words
    except:
        print('IndexError')

#视频地址
video_filename = r'本地视频位置'
videoCap = cv.VideoCapture(video_filename)
# 帧频
fps = videoCap.get(cv.CAP_PROP_FPS)
# 视频总帧数
total_frames = int(videoCap.get(cv.CAP_PROP_FRAME_COUNT))
# 图像尺寸
image_size = (int(videoCap.get(cv.CAP_PROP_FRAME_HEIGHT)), int(videoCap.get(cv.CAP_PROP_FRAME_WIDTH)))
#计算视频的时间秒数
video_time = int(total_frames / fps)
print(video_time)
# print(fps)
# print(total_frames)
# print(image_size)
#获取某一秒的image

def get_word_from_vedio(fpsnow):

    #调用 read() 方法可以读取一帧的图片，当 sucess 为 True 时，读取成功，此时 frame 是图像像素矩阵，三维 numpy 矩阵。
    for i in range(fpsnow):
        sucess, frame = videoCap.read()

    #确定字幕范围
    im = frame[:, :, 0]
    im = im[950:1050, 100:1800]     # 确定字幕的范围，注意不同的视频文件剪切的索引值不同，im[950:1050, 100:1900]分别是高度950-1050范围，宽度100-1900范围，根据视频情况调整

    #在进行下一步分析前，需对图像做二值化处理。由于字幕的白色的，像素值为 255，将二值化阈值设为 220。
    thresh = 160
    _, im = cv.threshold(im, thresh, 255, cv.THRESH_BINARY)
    img = Image.fromarray(im)
    # img.show()
    return im,img

#公式计算两张图像间每个像素点的平方误差之和的平均值百分比,当只大于1时出现字幕
def fps_percentage(im):
    fps_percentage_value = (im ** 2).sum() / im.size * 100
    if fps_percentage_value > 1:
        return True
    else:
        return False

#计算两张图差异大小值，误差值 > 1 时，字幕发生切换
def im_percentage(im1,im2):
    im_percentage = ((im2 - im1) ** 2).sum() / im1.size * 100
    if im_percentage > 1:
        return True
    else:
        return False

def main_work():
    for i in range(total_frames):
        i +=1
        im1,img1 = get_word_from_vedio(i)
        im2,img2 = get_word_from_vedio(i+2)
        vlue1 = fps_percentage(im1)
        vlue2 = fps_percentage(im2)
        if vlue1 and vlue2:
            print('--------\n都有文字')
            is_change_word = im_percentage(im1,im2)
            if is_change_word:
                print('文字发生变化')
                temp_pic1 = r'C:\Users\Administrator\desktop\test1.jpg'
                temp_pic2 = r'C:\Users\Administrator\desktop\test2.jpg'
                img1.save(temp_pic1)
                img2.save(temp_pic2)
                time.sleep(1)
                words1 = words_baidu_read(temp_pic1)
                words2 = words_baidu_read(temp_pic2)
                print('识别文字结果',words1,words2)
                if words1 in vedio_contents:
                    pass
                else:
                    vedio_contents.append([i,words1])
                if words2 in vedio_contents:
                    pass
                else:
                    vedio_contents.append([i,words2])

            else:
                print('文字没有变化')
            print('--------')
            pass
        elif vlue1:
            print('--------\n第1张文字，第2张没有')
            temp_pic1 = r'C:\Users\Administrator\desktop\test1.jpg'
            img1.save(temp_pic1)
            time.sleep(5)
            words1 = words_baidu_read(temp_pic1)
            if words1 in vedio_contents:
                pass
            else:
                vedio_contents.append([i,words1])
            print('--------')
            print('识别文字结果',words1)
            pass
        elif vlue2:
            print('--------\n第2张文字，第1张没有')
            temp_pic2 = r'C:\Users\Administrator\desktop\test2.jpg'
            img2.save(temp_pic2)
            time.sleep(5)
            words2 = words_baidu_read(temp_pic2)
            if words2 in vedio_contents:
                pass
            else:
                vedio_contents.append([i,words2])
            print('--------')
            print('识别文字结果',words2)
            pass
        else:
            pass

main_work()

#打印最后获得的list
for i in vedio_contents:
    print(i)