【图像处理】如何获取30FPS的1080P视频流

使用Logitech摄像头进行视频数据采集，发现在OpenCV中只能获取到5FPS的1080P数据。如果设置为30FPS，所获取到的分辨率会自动修改为640*480。原因在于摄像头默认的编码方式。

1.获取当前设备

$ lsusb
$ ls /dev/video*

下载luvcview，安装

sudo dpkg -i *.deb

查看设备信息

luvcview -d /dev/video0 -L

输出格式

Device information:
  Device path:  /dev/video0
{ pixelformat = 'YUYV', description = 'YUYV 4:2:2' }
{ discrete: width = 640, height = 480 }
    Time interval between frame: 1/30, 1/24, 1/20, 1/15, 1/10, 2/15, 1/5, 
{ discrete: width = 160, height = 90 }
    Time interval between frame: 1/30, 1/24, 1/20, 1/15, 1/10, 2/15, 1/5, 
...

表示在不同的视频格式下面，帧于帧之间的时间差。

2.测试

直接将获取到的视频分辨率设置为1920*1080，测试时间。

import cv2
import time

cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)

while True:
    time1 = time.time()
    ret, frame = cap.read()
    time2 = time.time()

    print('Frame time is: ', time2 - time1)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

测试结果为：

Frame time is:  0.1978147029876709
Frame time is:  0.20276236534118652
Frame time is:  0.1972665786743164
Frame time is:  0.19614434242248535
Frame time is:  0.19997501373291016
...

大约是每秒钟五帧。

修改编码格式（使用OpenCV的Fourcc）

import cv2
import time

cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)

cap.set(cv2.CAP_PROP_FOURCC, cv2.VideoWriter.fourcc('M', 'J', 'P', 'G'))
while True:
    time1 = time.time()
    ret, frame = cap.read()
    time2 = time.time()

    print('Frame time is: ', time2 - time1)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

再次测试速度：

Frame time is:  0.03886866569519043
Frame time is:  0.027163267135620117
Frame time is:  0.0348970890045166
Frame time is:  0.040320634841918945
Frame time is:  0.03423023223876953
Frame time is:  0.02848196029663086
Frame time is:  0.04337716102600098
Frame time is:  0.027630090713500977
Frame time is:  0.022168397903442383
...