【视频数据采集及前期处理（二）之视频数据前处理】

接上篇文章，得到的视频数据，需要前期处理，用于机器学习。功能目的：将视频数据去除画面中无人脸的帧，再另存为：

实现逻辑：通过opencv将视频流导入，取得每一帧进行人脸检测，if有人脸则将此帧数据存入：

#----------------------------------截取有人脸的视频——————————————————————————————
# -*- coding: utf-8 -*-
# import openCV的库
import cv2
import os, math, operator
from PIL import Image
from functools import reduce

# 读取输入视频流
video = cv2.VideoCapture("D:\\a000\\00NBD_data\\download\\s00001\\video\\s00001_001_00001.mp4")
# 获取视频的帧率和尺寸
fps = video.get(cv2.CAP_PROP_FPS)
width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
print(fps)
#####################################################################
# 创建输出VideoWriter对象以保存提取的帧为新的视频文件
output_video = cv2.VideoWriter(
    'img/video2.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))

# 告诉OpenCV使用人脸识别分类器(cv2.data.haarcascades可以用作数据文件夹的快捷方式)
classfier = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

#开始读取输入的视频帧
while video.isOpened():
    ok, frame = video.read()  # 读取一帧数据
    
    if not ok:
        break

    grey = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)  # 将当前桢图像转换成灰度图像

    # 人脸检测，1.2和2分别为图片缩放比例和需要检测的有效点数
    faceRects = classfier.detectMultiScale(grey, scaleFactor=1.2, minNeighbors=3, minSize=(32, 32))
    if len(faceRects) > 0:  # 大于0则检测到人脸

        output_video.write(frame)
    
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# 释放资源
video.release()
output_video.release()
cv2.destroyAllWindows()

发现有的间隔较大，删掉的部分较多，视频不连贯，没必要保存在同一段视频中，接下来截成几段：

#--------------------将视频有人脸的部分截下来，且间隔较长帧时间（这里设的5fps）未出现人脸，则分割存在下一段视频中
# -*- coding: utf-8 -*-
# import openCV的库
import cv2
import os, math, operator
from PIL import Image
from functools import reduce

# 读取输入视频流
video = cv2.VideoCapture("D:\\a000\\00NBD_data\\download\\s00001\\video\\s00001_001_00001.mp4")
# 获取视频的帧率和尺寸
fps = video.get(cv2.CAP_PROP_FPS)
width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
print(fps)
#####################################################################
# 创建输出VideoWriter对象以保存提取的帧为新的视频文件
#output_video = cv2.VideoWriter(
#    'img/video2.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))

# 告诉OpenCV使用人脸识别分类器(cv2.data.haarcascades可以用作数据文件夹的快捷方式)
classfier = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")
i = 1
count = 1
count_face = 0
output_video = cv2.VideoWriter(f'img/{i}.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))
#开始读取输入的视频帧
while video.isOpened():
    ok, frame = video.read()  # 读取一帧数据

    if not ok:
        break

    grey = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)  # 将当前桢图像转换成灰度图像

    # 人脸检测，1.2和2分别为图片缩放比例和需要检测的有效点数
    faceRects = classfier.detectMultiScale(grey, scaleFactor=1.2, minNeighbors=3, minSize=(32, 32))
    if len(faceRects) > 0:  # 大于0则检测到人脸

        if count-count_face < 5:

            output_video.write(frame)
        else:
            output_video.release()
            i+=1
            output_video = cv2.VideoWriter(f'img/{i}.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))

        count_face = count
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
    count += 1
# 释放资源
video.release()
cv2.destroyAllWindows()