写在前面
- 工作需要，简单整理
- 实际上这种方式延迟太高了，后来前端直接接的海康的本地解码插件，走的
websockt - 博文内容为
摄像头 rtsp 实时流转 hls m3u8的一个 Python 脚本 - 理解不足小伙伴帮忙指正 :),生活加油
99%的焦虑都来自于虚度时间和没有好好做事，所以唯一的解决办法就是行动起来，认真做完事情，战胜焦虑，战胜那些心里空荡荡的时刻，而不是选择逃避。不要站在原地想象困难，行动永远是改变现状的最佳方式
摄像头 rtsp 实时流转 hls m3u8 格式 web 端播放
方案介绍：
- 在服务器上安装并配置
FFmpeg,从RTSP摄像头获取实时视频流 - 使用
FFmpeg并将其转码为HLS格式,生成m3u8播放列表和TS分段文件。 - 将生成的
HLS文件托管到Nginx服务器的Web根目录下,并在Nginx配置文件中添加相应的配置,以正确处理 HLS 文件的 MIME 类型和跨域访问等。 - 在 Web 页面中使用 HTML5 的
<video>标签或HLS.js库来播放Nginx托管的HLS视频流。
这里使用的 Nginx 是有 rtmp 模块的 nginx https://github.com/dreammaker97/nginx-rtmp-win32-dev
rtsp 常见的两个转码方式：
rtsp 转 rtmp ffmpeg rtsp 2 rtmp
ffmpeg.exe -i rtsp://admin:hik12345@10.112.205.103:554/Streaming/Channels/101?transportmode=multicast -acodec aac -strict experimental -ar 44100 -ac 2 -b:a 96k -r 25 -b:v 500k -c:v libx264 -c:a copy -f flv rtmp://127.0.0.1:1935/live/demo
ffmpeg rtsp 2 hls rtsp 转 hls
ffmpeg -f rtsp -rtsp_transport tcp -i rtsp://admin:hik12345@10.112.205.103:554/Streaming/Channels/101?transportmode=multicast -acodec aac -strict experimental -ar 44100 -ac 2 -b:a 96k -r 25 -b:v 500k -s 640*480 -c:v libx264 -c:a copy -cpu-used 0 -threads 1 -f hls -hls_time 2.0 -hls_list_size 3 -hls_wrap 50 X:nginx-rtmp-win32-devnginx-rtmp-win32-devhtmlhlstest777.m3u8
名词解释:
RTSP 协议: RTSP (Real-Time Streaming Protocol) 是一种用于实时音视频流传输的网络协议,通常用于监控摄像头等设备的实时视频流传输。
HLS 格式: HLS (HTTP Live Streaming) 是苹果公司开发的自适应比特率流式传输协议,可以将视频流转码为 HTTP 可访问的 TS 分段文件和 m3u8 播放列表。HLS 具有良好的跨平台和兼容性。
FFmpeg : FFmpeg 是一个强大的多媒体框架,可以用于音视频的编码、解码、转码等操作。它可以将 RTSP 流转码为 HLS 格式。
Nginx: Nginx 是一款高性能的 Web 服务器,也可作为反向代理服务器使用。它可以托管 HLS 格式的 m3u8 播放列表和 TS 分段文件,为 Web 端提供 HLS 流的访问。
HLS.js: HLS.js 是一款 JavaScript 库,可以在不支持 HLS 原生播放的浏览器上实现 HLS 流的播放。
编码
通过 fastapi 启了一个Web服务，前端获取某个摄像头的流的时候，会启动一个 ffmpeg 子进程来处理流，同时会给前端返回一个 Nginx 推流的 地址
逻辑比较简单,涉及到进程处理，项目启动会自动启动 nginx，当取流时会自动启动 ffmpeg，nginx 和 ffmpge 都为 当前 Python 服务的子进程，当web 服务死掉，对应子进程全部死掉。
项目地址： https://github.com/LIRUILONGS/rtsp2hls-M3U8.git
requirements.txt
代码语言：javascript复制APScheduler==3.10.4
fastapi==0.111.1
ping3==4.0.8
pyinstaller==6.9.0
pytest==8.3.1
traitlets==5.14.3
uvicorn==0.30.3
配置文件
代码语言：javascript复制
# windows 环境配置文件，目录需要修改为 `/` 分割符
ngxin:
# 启动的推流服务IP，取流的时候使用的IP地址
nginx_ip : 127.0.0.1
# 启动 ng 端口，取流时使用的端口
nginx_port: 8080
# 启动的推流服务前缀
nginx_fix : /hls/
# nginx 程序路径，这里不加 `nginx.exe` 实际执行需要跳转到这个目录
nginx_path: "X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/"
# nginx 配置文件位置
nginx_config_path: "X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/conf/nginx.conf"
fastapi:
# 服务端口
port: 8991
# 流存放nginx目录
hls_dir: "X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/html/hls/"
# ffmpeg 执行路径
ffmpeg_dir: 'W:/ffmpeg-20200831-4a11a6f-win64-static/bin/ffmpeg.exe'
# 最大取流时间
max_stream_threads : 60
# 扫描时间
max_scan_time : 3*60
# 最大转码数
max_code_ff_size : 6
# ffmpeg 转化执行的路径
comm: "{ffmpeg_dir} -f rtsp -rtsp_transport tcp -i rtsp://admin:hik12345@{ip}:554/Streaming/Channels/101?transportmode=multicast -acodec aac -strict experimental -ar 44100 -ac 2 -b:a 96k -r 25 -b:v 300k -s {width}*{height} -live 1 -c:v libx264 -c:a copy -cpu-used 0 -threads 1 -f hls -hls_time 0.5 -hls_list_size 1 -hls_wrap 100 {hls_dir}{ip}-{uuid_v}.m3u8"
核心代码
代码语言：javascript复制#!/usr/bin/env python
# -*- encoding: utf-8 -*-
"""
@File : main.py
@Time : 2024/07/24 17:20:21
@Author : Li Ruilong
@Version : 1.0
@Contact : liruilonger@gmail.com
@Desc : rtmp 转码 到 hls
"""
........................................
@app.get("/sc_view/get_video_stream")
async def get_video_stream(
ip: str = Query("192.168.2.25", description="IP地址"), # 设置默认值为 1
width: int = Query(320, description=" 流宽度"), # 设置默认值为 10
height: int = Query(170, description=" 流高度"), # 设置默认值为 'name'
):
"""
@Time : 2024/07/23 11:04:31
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : ffmag 解码推流
"""
if width is None or ip is None or height is None:
raise HTTPException(status_code=400, detail="参数不能为空")
import time
# 获取前端传递的参数
uuid_v = str(uuid.uuid4())
if validate_ip_address(ip) is False:
return {"message": "no validate_ip_address", "code": 600}
if ping_test(ip) is False:
return {"message": "ping no pong", "code": 600}
with lock:
if ip in chanle:
return chanle[ip]
if len(chanle) >= max_code_ff_size:
return {"status": 400, "message": f"超过最大取流数：{max_code_ff_size}"}
hls_dir = fastapi['hls_dir']
ffmpeg_dir = fastapi["ffmpeg_dir"]
print(vars())
command = comm.format_map(vars())
try:
print(command.strip())
process = subprocess.Popen(
command,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL
)
if process.pid:
t_d = {
"pid": process.pid,
"v_url": f'http://{locad_id}:{locad_port}{locad_fix}{ip}-{uuid_v}.m3u8',
"ip": ip
}
print(t_d)
print("==============================摄像头数据更新完成...，重新确认子进程是否运行")
pss = get_process_by_name("ffmpeg.exe", process.pid)
print("创建的进程为:", pss)
if len(pss) > 0:
chanle[ip] = t_d
print(f"返回取流路径为：{t_d}")
return t_d
else:
return {"status": 400, "message": "IP 取流失败!,请重新尝试"}
except subprocess.CalledProcessError as e:
return {"error": f"Error running ffmpeg: {e}"}
@app.get("/sc_view/stop_video_stream")
async def stop_video_stream(pid: int = Query(2000, description="进程ID")):
"""
@Time : 2024/07/24 14:10:43
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : 结束推流
"""
if pid is None:
raise HTTPException(status_code=400, detail="参数不能为空")
pss = get_process_by_name("ffmpeg.exe", pid)
print(pss)
if len(pss) == 0:
print("未获取到进程信息", pid)
return {
"status": 200,
"message": "未获取到进程信息"
}
print("获取到进程信息：", pss)
try:
# 发送 SIGTERM 信号以关闭进程
os.kill(int(pid), signal.SIGTERM)
chanle.pop(pid)
print(f"Process {pid} has been terminated.{str(pss)}")
return {"status": 200, "message": "关闭成功!"}
except OSError as e:
# 调用 kill 命令杀掉
pss[0].kill()
print(f"Error terminating process {pid}: {e}")
return {"status": 200, "message": "关闭成功!"}
@app.get("/sc_view/all_stop_video_stream")
async def all_stop_video_stream():
"""
@Time : 2024/07/24 14:10:43
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : 批量结束推流
"""
pss = get_process_by_name("ffmpeg.exe")
print(pss)
if len(pss) == 0:
return {
"status": 200,
"message": "转码全部结束"
}
print("获取到进程信息：", pss)
process_list = []
for p in pss:
process_list.append({
"pid": p.info['pid'],
"name": p.info['name'],
"status": p.status(),
"started": datetime.datetime.fromtimestamp(p.info['create_time']),
"memory_info": str(p.memory_info().rss / (1024 * 1024)) " MB",
"cpu_percent": str(p.cpu_percent()) " %",
"cmdline": p.cmdline()
})
try:
# 发送 SIGTERM 信号以关闭进程
os.kill(int(p.info['pid']), signal.SIGTERM)
#chanle.pop(p.info['pid'])
ips = [ k for k,v in chanle.items() if v.pid == p.info['pid'] ]
if len(ips) >0:
chanle.pop(ips[0])
print(f"Process {p.info['pid']} has been terminated.{str(pss)}")
except OSError as e:
# 调用 kill 命令杀掉
pss[0].kill()
print(f"Error terminating process {p.info['pid']}: {e}")
return {"status": 200, "message": "关闭成功!", "close_list": process_list}
@app.get("/sc_view/get_video_stream_process_list")
async def get_video_stream_process_list():
"""
@Time : 2024/07/24 15:46:38
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : 返回当前在采集的流处理进程信息
"""
pss = get_process_by_name("ffmpeg.exe")
process_list = []
for p in pss:
ip_file = str(p.info['cmdline'][-1]).split("/")[-1]
process_list.append({
"pid": p.info['pid'],
"name": p.info['name'],
"status": p.status(),
"started": datetime.datetime.fromtimestamp(p.info['create_time']),
"memory_info": str(p.memory_info().rss / (1024 * 1024)) " MB",
"cpu_percent": str(p.cpu_percent()) " %",
"cmdline": p.cmdline(),
"v_url": f'http://{locad_id}:{locad_port}{locad_fix}{ip_file}',
})
return {"message": "当前在采集的流信息", "process_list": process_list}
nginx 启动相关
代码语言：javascript复制# 启动 Nginx
def start_nginx():
"""
@Time : 2024/07/24 21:13:25
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : 启动 nginx
"""
try:
os.chdir(nginx_path)
print("当前执行路径：" str(nginx_path "nginx.exe" " -c " nginx_config_path))
subprocess.Popen([nginx_path "nginx.exe", "-c", nginx_config_path], stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL)
print("n=================== Nginx has been started successfully.n")
except subprocess.CalledProcessError as e:
print(f"Failed to start Nginx: {e}")
finally:
os.chdir(os.path.dirname(__file__)) # 切换回用户主目录
# 停止 Nginx
def stop_nginx():
"""
@Time : 2024/07/24 21:13:41
@Author : liruilonger@gmail.com
@Version : 1.0
@Desc : 关闭 nginx
"""
try:
os.chdir(nginx_path)
print("当前执行路径：" str(nginx_path "nginx.exe" " -s " "stop"))
subprocess.Popen([nginx_path "nginx.exe", "-s", "stop"], stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL)
print("n============ Nginx has been stopped successfully.n")
except subprocess.CalledProcessError as e:
print(f"Failed to stop Nginx: {e}")
finally:
os.chdir(os.path.dirname(__file__)) # 切换回用户主目录
进程相关方法
代码语言：javascript复制def get_process_by_name(process_name, pid=None):
"""
@Time : 2024/07/24 14:21:31
@Author : liruilonger@gmail.com
@Version : 1.1
@Desc : 获取指定进程名和进程 ID 的进程列表
Args:
process_name (str): 进程名称
pid (int, optional): 进程 ID，默认为 None 表示不筛选 ID
Returns:
list: 包含指定进程名和进程 ID 的进程对象的列表
"""
processes = []
attrs = ['pid', 'memory_percent', 'name', 'cmdline', 'cpu_times',
'create_time', 'memory_info', 'status', 'nice', 'username']
for proc in psutil.process_iter(attrs):
# print(proc.info['name'])
try:
if proc.info['name'] == process_name:
if pid is None or proc.info['pid'] == pid:
processes.append(proc)
except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess):
pass
print("Process==================end")
return processes
def get_process_by_IP(process_name, ip=None):
"""
@Time : 2024/07/24 14:21:31
@Author : liruilonger@gmail.com
@Version : 1.1
@Desc : 获取指定进程名和 IP 的进程列表
Args:
process_name (str): 进程名称
pid (int, optional): IP，默认为 None 表示不筛选 IP
Returns:
list: 包含指定进程名和进程 IP 的进程对象的列表
"""
attrs = ['pid', 'memory_percent', 'name', 'cmdline', 'cpu_times',
'create_time', 'memory_info', 'status', 'nice', 'username']
press = []
for proc in psutil.process_iter(attrs):
try:
if proc.info['name'] == process_name:
if ip is None or any(ip in s for s in proc.info['cmdline']):
ip_file = str(proc.info['cmdline'][-1]).split("/")[-1]
press.append({
"pid": proc.info['pid'],
"v_url": f'http://{locad_id}:{locad_port}{locad_fix}{ip_file}',
"ip": ip
})
except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess):
pass
return press
打包
代码语言：javascript复制pyinstaller --add-data "config.yaml;." --add-data "templates/*;templates" main.py
exe 路径
代码语言：javascript复制rtsp2hls2M3U8distmain
配置文件路径
代码语言：javascript复制rtsp2hls2M3U8distmain_internal
部署测试
代码语言：javascript复制2024-08-13 15:57:03,404 - win32.py[line:58] - DEBUG: Looking up time zone info from registry
2024-08-13 15:57:03,410 - yaml_util.py[line:62] - INFO: 加载配置数据：{'ngxin': {'nginx_ip': '127.0.0.1', 'nginx_port': 8080, 'nginx_fix': '/hls/', 'nginx_path': 'X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/', 'nginx_config_path': 'X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/conf/nginx.conf'}, 'fastapi': {'port': 8991, 'hls_dir': 'X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/html/hls/', 'ffmpeg_dir': 'W:/ffmpeg-20200831-4a11a6f-win64-static//bin/ffmpeg.exe', 'max_stream_threads': 60, 'max_scan_time': '3*60', 'max_code_ff_size': 6, 'comm': '{ffmpeg_dir} -f rtsp -rtsp_transport tcp -i rtsp://admin:hik12345@{ip}:554/Streaming/Channels/101?transportmode=multicast -acodec aac -strict experimental -ar 44100 -ac 2 -b:a 96k -r 25 -b:v 300k -s {width}*{height} -live 1 -c:v libx264 -c:a copy -cpu-used 0 -threads 1 -f hls -hls_time 0.5 -hls_list_size 1 -hls_wrap 100 {hls_dir}{ip}-{uuid_v}.m3u8'}}
2024-08-13 15:57:03,413 - base.py[line:454] - INFO: Adding job tentatively -- it will be properly scheduled when the scheduler starts
2024-08-13 15:57:03,414 - proactor_events.py[line:630] - DEBUG: Using proactor: IocpProactor
INFO: Started server process [30404]
INFO: Waiting for application startup.
2024-08-13 15:57:03,441 - base.py[line:895] - INFO: Added job "scan_video_stream_list" to job store "default"
2024-08-13 15:57:03,441 - base.py[line:181] - INFO: Scheduler started
Process==================end
当前执行路径：X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/nginx.exe -c X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/conf/nginx.conf
=================== Nginx has been started successfully.
2024-08-13 15:57:09,256 - base.py[line:954] - DEBUG: Looking for jobs to run
2024-08-13 15:57:09,256 - base.py[line:1034] - DEBUG: Next wakeup is due at 2024-08-13 16:57:03.413311 08:00 (in 3594.156780 seconds)
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL C to quit)
API 文档:http://127.0.0.1:8000/docs#
测试页面
代码语言：javascript复制{'ip': '192.168.2.25', 'width': 320, 'height': 170, 'time': <module 'time' (built-in)>, 'uuid_v': 'dbeda9ce-01ec-41cd-8315-8145954d1ea0', 'hls_dir': 'X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/html/hls/', 'ffmpeg_dir': 'W:/ffmpeg-20200831-4a11a6f-win64-static//bin/ffmpeg.exe'}
W:/ffmpeg-20200831-4a11a6f-win64-static//bin/ffmpeg.exe -f rtsp -rtsp_transport tcp -i rtsp://admin:hik12345@192.168.2.25:554/Streaming/Channels/101?transportmode=multicast -acodec aac -strict experimental -ar 44100 -ac 2 -b:a 96k -r 25 -b:v 300k -s 320*170 -live 1 -c:v libx264 -c:a copy -cpu-used 0 -threads 1 -f hls -hls_time 0.5 -hls_list_size 1 -hls_wrap 100 X:/nginx-rtmp-win32-dev/nginx-rtmp-win32-dev/html/hls/192.168.2.25-dbeda9ce-01ec-41cd-8315-8145954d1ea0.m3u8
{'pid': 32416, 'v_url': 'http://127.0.0.1:8080/hls/192.168.2.25-dbeda9ce-01ec-41cd-8315-8145954d1ea0.m3u8', 'ip': '192.168.2.25'}
==============================摄像头数据更新完成...，重新确认子进程是否运行
Process==================end
创建的进程为: [psutil.Process(pid=32416, name='ffmpeg.exe', status='running', started='15:59:38')]
返回取流路径为：{'pid': 32416, 'v_url': 'http://127.0.0.1:8080/hls/192.168.2.25-dbeda9ce-01ec-41cd-8315-8145954d1ea0.m3u8', 'ip': '192.168.2.25'}
INFO: 127.0.0.1:64650 - "GET /sc_view/get_video_stream?ip=192.168.2.25&width=320&height=170 HTTP/1.1" 200 OK
博文部分内容参考
© 文中涉及参考链接内容版权归原作者所有，如有侵权请告知 :)
© 2018-2024 liruilonger@gmail.com, 保持署名-非商用-相同方式共享(CC BY-NC-SA 4.0)
