最近做的一个项目需要使用arm开发板调用摄像头获取图片数据发送到后端,由于开发板是fedora系统,所以踩了不少坑。特此记录下来。
https://developer.aliyun.com/mirror/fedora
https://blog.csdn.net/who__are__you_/article/details/82628390
$ dnf install xxx
$ Error: Failed to synchronize cache for repo
https://blog.csdn.net/l1175832366/article/details/104175396
执行以下命令:
sed -i 's/releases\/$releasever/releases\/32/g' /etc/yum.repos.d/fedora.repo
sed -i 's/updates\/$releasever/updates\/32/g' /etc/yum.repos.d/fedora-updates.repo
https://blog.csdn.net/heng615975867/article/details/80519274
配置完成后,把下载目录里面的nginx-rtmp-module-master移到/usr/local/nginx中,因为可能没有移过去
https://www.cnblogs.com/codeAB/p/9167107.html
安装完之后可能需要添加环境变量:
vim /etc/profile
添加一行
export PATH=/usr/local/ffmpeg/bin/:$PATH\
刷新环境
souce /etc/profile
如果在命令行中输入ffmpeg没有报找不到的错误就成功。如果显示找不到,那就是环境变量没有配置好。
将里面的地址修改成你要推流的地址
rtmpUrl = "rtmp://139.159.142.192:1935/live/1"
如果想要调用摄像头的参数设为0显示找不到摄像头,就参考下面的查找本机摄像头序号。因为我用的是AI开发板,所以直接输入0可能调用不了。
vid = cv2.VideoCapture(r"/usr/local/web/studey/mysite/chat/video/4.mp4")
import cv2
import queue
import os
import numpy as np
from threading import Thread
import datetime, _thread
import subprocess as sp
from time import *
# 使用线程锁,防止线程死锁
mutex = _thread.allocate_lock()
# 存图片的队列
frame_queue = queue.Queue()
# 推流的地址,前端通过这个地址拉流,主机的IP,2019是ffmpeg在nginx中设置的端口号
rtmpUrl = "rtmp://139.159.142.192:1935/live/1"
# 用于推流的配置,参数比较多,可网上查询理解
command = ['ffmpeg',
'-y',
'-f', 'rawvideo',
'-vcodec', 'rawvideo',
'-pix_fmt', 'bgr24',
'-s', "{}x{}".format(640, 480), # 图片分辨率
'-r', str(25.0), # 视频帧率
'-i', '-',
'-c:v', 'libx264',
'-pix_fmt', 'yuv420p',
'-preset', 'ultrafast',
'-f', 'flv',
rtmpUrl]
def Video():
# 调用相机拍图的函数
vid = cv2.VideoCapture(r"/usr/local/web/studey/mysite/chat/video/4.mp4")
if not vid.isOpened():
raise IOError("Couldn't open webcam or video")
while (vid.isOpened()):
return_value, frame = vid.read()
# 原始图片推入队列中
frame_queue.put(frame)
def push_frame():
# 推流函数
accum_time = 0
curr_fps = 0
fps = "FPS: ??"
prev_time = time()
# 防止多线程时 command 未被设置
while True:
if len(command) > 0:
# 管道配置,其中用到管道
p = sp.Popen(command, stdin=sp.PIPE)
break
while True:
if frame_queue.empty() != True:
# 从队列中取出图片
frame = frame_queue.get()
# curr_time = timer()
# exec_time = curr_time - prev_time
# prev_time = curr_time
# accum_time = accum_time + exec_time
# curr_fps = curr_fps + 1
# process frame
# 你处理图片的代码
# 将图片从队列中取出来做处理,然后再通过管道推送到服务器上
# 增加画面帧率
# if accum_time > 1:
# accum_time = accum_time - 1
# fps = "FPS: " + str(curr_fps)
# curr_fps = 0
# write to pipe
# 将处理后的图片通过管道推送到服务器上,image是处理后的图片
p.stdin.write(frame.tostring())
def run():
# 使用两个线程处理
thread1 = Thread(target=Video, )
thread1.start()
thread2 = Thread(target=push_frame, )
thread2.start()
if __name__ == '__main__':
run()
拉流使用你的推流地址
vid = cv2.VideoCapture("你的推流地址")
https://blog.csdn.net/weixin_44345862/article/details/91047938?ops_request_misc=&request_id=&biz_id=102&utm_term=%E6%9F%A5%E7%9C%8B%E8%BF%9E%E6%8E%A5%E6%91%84%E5%83%8F%E5%A4%B4%E5%8F%B7&utm_medium=distribute.pc_search_result.none-task-blog-2allsobaiduweb~default-0-91047938.nonecase&spm=1018.2226.3001.4187
https://blog.csdn.net/keith_bb/article/details/54172899?spm=1001.2101.3001.6650.2&utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7EOPENSEARCH%7Edefault-2.tagcolumn&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2%7Edefault%7EOPENSEARCH%7Edefault-2.tagcolumn