0 前言

视频是按照一定规则组织的图像序列，我们按照规则获取即可

1 对取视频对象，并获取基本信息

cap = cv2.VideoCapture(video_filepath)
# 读取视频的基本信息
# 获取FPS
fps = int(cap.get(cv2.CAP_PROP_FPS))
# 获取总共帧的个数
total_slice = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
# 获取视频的宽和高
height, width = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)), int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))

Note: 获取的fps 与 total_slice 为浮点数，这里转化为整型

2 按照fps读取视频帧

import cv2
import numpy as npdef extract_frame(video_filepath, save_image_filepath, start_time, end_time, clip_num=1, log=False, dsize=None, image_count=0, direction=1):cap = cv2.VideoCapture(video_filepath)# 读取视频的基本信息# 获取FPSfps = int(cap.get(cv2.CAP_PROP_FPS))# 获取总共帧的个数total_slice = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))# 获取视频的宽和高height, width = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)), int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))if log:print(f"video info: fps: {fps}, slice sum: {total_slice}, height: {height}, width: {width}")images = []frames = [fps * clip_num * j + image_count for j in range(start_time, end_time)]if cap.isOpened():for i in range(int(total_slice)):ret, frame = cap.read()image = frameif dsize is not None:image = cv2.resize(src=frame, dsize=dsize, interpolation=cv2.INTER_LINEAR)if i in frames:images.append(image)# 释放内存cap.release()images = np.concatenate(images, axis=direction)cv2.imwrite(filename=save_image_filepath, img=images)if __name__ == '__main__':video_filepath = "/sharefiles1/hanliqiang/pythonProject/Pyhton_movie/CTRL.mp4"save_image_filepath = "./test_extract_images.png"extract_frame(video_filepath=video_filepath, save_image_filepath=save_image_filepath, start_time=15,end_time=25)pass

• 这里只读取了每一帧的第一张图像，可以通过设置image_count=15来获取中间帧
• 对图像进行resize，指定图像大小即可，这里为tuple类型
• direction=0 水平拼接， direction=1 垂直拼接
• clip_num 为间隔帧

3 Experiment Result

• 视频资料请自行下载 https://www.bilibili.com/video/BV1td4y1T7CB

• clip_num = 2

• clip_num=1

参考

python实现读取和保存视频: https://blog.csdn.net/weixin_52777510/article/details/119872183