Opencv读取视频获取的总帧数比实际少很多问题

Opencv读取视频获取的总帧数比实际少很多问题

问题描述

​ 在读取一段从摄像头拉流保存下来的8分钟视频，opencv(opencv4.4)获取到的总帧数只有7帧，循环读取解码视频，获取到的帧数也只有7帧，明显此时存在问题

产生问题背景

​ opencv读取拉流视频，可能不同的解码器，会导致在调用opencv自身的解码器无法对视频进行解码，导致一部分帧解码失败，导致实际看到的帧数和读取到的帧数相差甚远

opencv视频解码code

import cv2
video_file = "Camera.mp4"
capture = cv2.VideoCapture(video_file)      #读取视频或摄像               

# Get Video info : resolution, fps, frame count
width = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH))         #宽
height = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))       #高
fps = int(capture.get(cv2.CAP_PROP_FPS))                   #帧率
frame_count = int(capture.get(cv2.CAP_PROP_FRAME_COUNT))   #总帧数
print("video fps: %d, frame_count: %d,width: %d,height: %d" % (fps, frame_count,width,height))
frameNr = 0
while True:
    ret, frame = capture.read() #opencv 读取到的视频帧默认是bgr
    if not ret:
        break
    # frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) #图片bgr转换成rgb
    cv2.imshow('demo', frame)
    if cv2.waitKey(30) & 0xFF == ord('q'):
        break
    frameNr = frameNr + 1
    
    
# 关闭视频文件
capture.release() 
cv2.destroyAllWindows()

video fps: 25, frame_count: 6,width: 1920,height: 1080

解决办法

方法一:对视频通过ffmpeg解码

ffmpeg -i Camera_0420230221104457_20230221111458Centralized Storage.mp4 out.mp4

ffmpeg相关教程

• ffmpeg常用编解码命令
• ffmpeg 码率控制
• ffmpeg 使用例程

方法二:使用imageio-ffmpeg读取视频，对视频进行解码

安装ffmpeg解码器

apt-get install ffmpeg

Centos安装ffmpeg

安装imageio-ffmpeg 包

pip install imageio-ffmpeg -i https://pypi.douban.com/simple

imageio视频解码code

import cv2
import numpy as np
import imageio

#pip install imageio-ffmpeg -i https://pypi.douban.com/simple

video_file = "Camera.mp4"
cap = imageio.get_reader(video_file, 'ffmpeg')   #读取视频或摄像   
fps = cap.get_meta_data()['fps']                 #帧率
(width,height) = cap.get_meta_data()['size']     #size,(宽，高)     
duration = cap.get_meta_data()['duration']       #时长
# frame_count = cap.count_frames()                #总帧率 可以通过时长x帧率获取(存在异常AttributeError: 'Reader' object has no attribute 'count_frames')
frame_count = int(fps*duration)                   #总帧率 可以通过时长x帧率获取
print("video fps: %d, frame_count: %d,width: %d,height: %d" % (fps, frame_count,width,height))

for frameNr,frame in enumerate(cap): #imageio 读取到的视频帧默认是rgb
    #image的类型是imageio.core.util.Image可用使用的下面这一注释行转换为arrary
    #image = skimage.img_as_float(im).astype(np.float32)
    frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    cv2.imshow('demo', frame_rgb)
    if cv2.waitKey(30) & 0xFF == ord('q'):
        break
    frameNr = frameNr + 1
     
# 关闭视频文件
cap.close()

video fps: 25, frame_count: 10814,width: 1920,height: 1080