使用OpenCV合并两个.AVI视频失败，求解决方案

问题描述 #

尝试用OpenCV将两个AVI视频按顺序合并，输出文件仅14KB且无法在VLC播放。两个输入视频由Fiji的TrackMate生成，大小分别为6.6MB和6.2MB，时长均为10秒。使用的代码如下：

import cv2

path = '/Volumes/GQA HD/GQA Experiments/2023_01_27/'
file1 = 'TrackMate capture of 1.avi'
file2 = 'TrackMate capture of 2.avi'

file_combined = 'TrackMate Combined captures of IMG_0240.avi'

# Paths of videos
videos = [path+file1, path+file2]

# Create new video
fourcc = cv2.VideoWriter_fourcc('M','J','P','G')
frameSize=(720,1280)
fps=30
video = cv2.VideoWriter(path+file_combined, fourcc, fps, frameSize)

# Write all frames sequentially to new video
for v in videos:
    curr_v = cv2.VideoCapture(v)
    while curr_v.isOpened():
        r, frame = curr_v.read()    # Get return value and curr frame of curr video
        if not r:
            break
        video.write(frame)          # Write the frame

video.release()                     # Save the video

解决方案 #

• 匹配原视频的关键参数
  手动指定的frameSize和fps可能与输入视频实际参数不匹配，导致VideoWriter无法正常写入帧。应该从输入视频中读取这些参数：

  # 从第一个视频获取基准参数
  cap = cv2.VideoCapture(videos[0])
  fps = cap.get(cv2.CAP_PROP_FPS)
  frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
  frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
  cap.release()
  
  # 用原视频参数初始化VideoWriter
  fourcc = cv2.VideoWriter_fourcc(*'MJPG')
  video = cv2.VideoWriter(path+file_combined, fourcc, fps, (frame_width, frame_height))
  

• 添加视频读取有效性检查
  确保每个输入视频都能成功打开，避免因读取失败导致空帧写入：

  for v in videos:
      curr_v = cv2.VideoCapture(v)
      if not curr_v.isOpened():
          print(f"无法加载视频文件: {v}")
          continue
      while curr_v.isOpened():
          ret, frame = curr_v.read()
          if not ret:
              break
          # 确保帧尺寸与输出一致，不一致则调整
          if frame.shape[:2] != (frame_height, frame_width):
              frame = cv2.resize(frame, (frame_width, frame_height))
          video.write(frame)
      curr_v.release()  # 及时释放当前视频资源
  

• 验证FourCC编码兼容性
  MJPG编码虽通用，但部分环境可能存在兼容性问题。如果仍失败，可尝试替换为cv2.VideoWriter_fourcc(*'XVID')，或者留空参数让OpenCV自动适配（部分版本支持）。

• 确保资源彻底释放
  每个输入视频处理完成后必须调用curr_v.release()，避免资源泄漏导致视频写入不完整。

内容的提问来源于stack exchange，提问作者Steven Basmith