如何用OpenCV读取Elgato HD60 Pro采集卡视频流？

问题 #

尝试用Python的OpenCV读取Elgato HD 60 Pro采集卡的视频流，但持续报错。已确认采集卡正常工作，Python和OpenCV均为最新版本，但程序无法识别设备。

首先尝试枚举设备ID，代码如下：

def find_capture_device_id(max_tests=10):
    for device_id in range(max_tests):
        capture = cv2.VideoCapture(device_id)
        if capture is None or not capture.isOpened():
            if capture is not None:
                capture.release()
            continue

        ret, frame = capture.read()
        capture.release()

        if ret:
            print(f"Device ID {device_id} is available and returns valid frames.")
        else:
            print(f"Device ID {device_id} is available but does not return valid frames.")

运行后，设备ID 0可被识别但无法获取有效帧，错误信息如下：

[ WARN:0@10.925] global cap_msmf.cpp:1768 CvCapture_MSMF::grabFrame videoio(MSMF): can't grab frame. Error: -2147483638
Device ID 0 is available but does not return valid frames.
[ERROR:0@11.664] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@11.819] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@11.968] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.117] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.278] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.441] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.593] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.752] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range
[ERROR:0@12.941] global obsensor_uvc_stream_channel.cpp:159 cv::obsensor::getStreamChannelGroup Camera index out of range

随后尝试用设备管理器获取的PCI ID指定设备，代码如下：

vendor_id = "12AB"
device_id = "0380"

# Combine them into the device ID string
device_id_string = f"PCI\\VEN_{vendor_id}&DEV_{device_id}"

# Try different capture methods
capture = cv2.VideoCapture(device_id_string)
if not capture.isOpened():
    # Try using cv2.CAP_DSHOW
    capture = cv2.VideoCapture(cv2.CAP_DSHOW)
    if not capture.isOpened():
        # Try using cv2.CAP_VFW
        capture = cv2.VideoCapture(cv2.CAP_VFW)
        if not capture.isOpened():
            print(f"Device ID {device_id_string} is not available or does not return valid frames.")
            return

print("Pass")

依旧失败，错误信息：

Device ID PCI\VEN_12AB&DEV_0380 is not available or does not return valid frames.
[ WARN:0@1.129] global cap.cpp:344 cv::VideoCapture::open VIDEOIO(DSHOW): backend is generally available but can't be used to capture by index

请问哪里操作有误？如何配置OpenCV或使用其他工具读取该采集卡的视频流？

解决方案与替代方案 #

一、OpenCV的正确调整 #

1. 指定MSMF后端并强制设置分辨率
   Elgato HD60 Pro是PCIe采集卡，OpenCV默认的MSMF后端需要明确匹配输入源的分辨率才能正常抓取帧。修改代码如下：

import cv2

def test_elgato_capture():
    # 指定MSMF后端打开设备0
    cap = cv2.VideoCapture(0, cv2.CAP_MSMF)
    # 设置与输入源一致的分辨率（示例为1080p 60fps，根据你的实际输入调整）
    cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
    cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
    cap.set(cv2.CAP_PROP_FPS, 60)
    
    if not cap.isOpened():
        print("无法打开采集卡设备")
        return
    
    while True:
        ret, frame = cap.read()
        if not ret:
            print("无法获取视频帧")
            break
        
        cv2.imshow("Elgato Capture", frame)
        # 按q键退出
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break
    
    cap.release()
    cv2.destroyAllWindows()

test_elgato_capture()

2. 纠正PCI ID的错误用法
   OpenCV的VideoCapture不支持直接传入PCI设备ID，你之前的写法完全错误。设备只能通过索引（0、1、2...）、视频文件路径或网络流URL来打开，PCI ID无法被OpenCV识别。

二、替代工具推荐 #

如果OpenCV仍无法正常工作，推荐以下兼容性更好的方案：

• OBS虚拟摄像头 + pyvirtualcam
  先通过OBS Studio捕获Elgato采集卡的流，开启OBS的虚拟摄像头功能，再用pyvirtualcam读取虚拟摄像头的流。OBS对采集卡的支持非常完善，几乎不会出现兼容性问题。
  示例代码：

import pyvirtualcam
import cv2

# 这里的0是OBS虚拟摄像头的设备索引，根据实际情况调整
cap = cv2.VideoCapture(0)
# 匹配OBS设置的分辨率和帧率
with pyvirtualcam.Camera(width=1920, height=1080, fps=60) as cam:
    while True:
        ret, frame = cap.read()
        if not ret:
            break
        # OpenCV默认是BGR格式，pyvirtualcam需要RGB格式
        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        cam.send(frame_rgb)
        cam.sleep_until_next_frame()

• Elgato官方Game Capture SDK
  直接使用Elgato官方提供的Game Capture SDK，这是兼容性和功能最优的方案，不过需要注册Elgato开发者账号获取SDK权限。

三、额外排查要点 #

• 确保Elgato官方驱动已安装，且采集卡未被其他程序占用（比如OBS、Streamlabs等）。
• 尝试以管理员身份运行Python脚本，部分采集卡需要系统权限才能访问。
• 若使用OpenCV 5.x版本，可尝试降级到4.x版本，新版本可能存在兼容性bug。

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