使用OpenCV提取视频帧时遭遇grabFrame packet read max attempts exceeded报错

问题描述 #

使用OpenCV提取视频帧、运行分割AI模型并保存帧与掩码至文件夹时，处理一定帧数后触发报错，多环境下同一视频稳定复现。已尝试调高OPENCV_FFMPEG_READ_ATTEMPTS至10000，无明显改善。

报错信息 #

[ WARN:0@379.898] global cap_ffmpeg_impl.hpp:1541 grabFrame packet read max attempts exceeded, if your video have multiple streams (video, audio) try to increase attempt limit by setting environment variable OPENCV_FFMPEG_READ_ATTEMPTS (current value is 10000)

最小复现代码 #

import os
import cv2

videofilename = "test.mp4"
capture = cv2.VideoCapture(videofilename)
frameNum = 0

createfolder = os.getcwd() + '/' + videofilename.split(".")[0] + '/'
if not os.path.exists(createfolder):
    os.makedirs(createfolder)
    os.makedirs(createfolder + "/frames/")

while True:
    success, frame = capture.read()
    if success is False:
        break
    frameNum += 1
    framedownloadname = videofilename.split(".")[0] + '-fr' + str(frameNum) + '.jpg'
    framedownloadloc = createfolder + '/frames/' + framedownloadname
    print(framedownloadloc)
    cv2.imwrite(framedownloadloc, frame)
    img = cv2.imread(framedownloadloc)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

capture.release()

解决方案 #

• 指定仅读取视频流：OpenCV默认会尝试读取所有流（视频+音频），若视频音频流存在损坏或索引异常，会导致读取卡住。初始化VideoCapture时指定仅读取视频流：

  capture = cv2.VideoCapture(videofilename, cv2.CAP_FFMPEG)
  # 设置仅解码视频流
  capture.set(cv2.CAP_PROP_STREAM_TYPE, cv2.CAP_STREAM_VIDEO)
  

• 跳过异常帧继续处理：修改循环逻辑，在读取失败时尝试重新抓取帧，跳过无效帧后继续：

  while True:
      success, frame = capture.read()
      if not success:
          # 尝试重新抓取帧
          grab_success = capture.grab()
          if not grab_success:
              break
          # 提取抓取到的帧
          success, frame = capture.retrieve()
          if not success:
              continue
      # 后续帧处理逻辑不变
      frameNum += 1
      framedownloadname = videofilename.split(".")[0] + '-fr' + str(frameNum) + '.jpg'
      framedownloadloc = createfolder + '/frames/' + framedownloadname
      print(framedownloadloc)
      cv2.imwrite(framedownloadloc, frame)
      img = cv2.imread(framedownloadloc)
      img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
  

• 修复视频文件索引：若视频本身存在编码或容器损坏，用FFmpeg工具修复视频索引后再处理：

  ffmpeg -i test.mp4 -c:v copy -c:a copy -fix_sub_duration output_fixed.mp4
  

• 调整FFmpeg读取参数：进一步调高读取尝试次数，并设置FFmpeg读取超时，在初始化VideoCapture前配置环境变量：

  import os
  os.environ['OPENCV_FFMPEG_READ_ATTEMPTS'] = '50000'
  # 设置TCP传输及超时时间（单位微秒）
  os.environ['OPENCV_FFMPEG_CAPTURE_OPTIONS'] = 'rtsp_transport;tcp|timeout;5000000'
  capture = cv2.VideoCapture(videofilename, cv2.CAP_FFMPEG)
  

内容的提问来源于stack exchange，提问作者banjaxing