从 C++ 调用网络摄像头并在 Python 中显示图像

Question

I want to call a function from C++ in Python.

In my case, I already have a lot of code in Python but I want to change the camera that I use. Some libraries for cameras only support C++.

I have successfully call C++.dll and displaying images in Python with my code as follow. However, when Python show images very slow.

I think it becomes slow because I need to call VideoCapture in a loop. If I only enter VideoCapture 1 time (before main function), the image only send 1 time also in Python.

This is my code for C++

#include <iostream>
#include <fstream>
#include <sstream>
#include <stdio.h>
#include <vector>
#include <opencv2\opencv.hpp>
#include <opencv2\highgui\highgui.hpp>

using namespace std;
using namespace cv;

Mat img;
Mat frameRGB;
vector<Mat>ch;
Mat1f datas, datas2, datas3, datas4;
float tmp_buf[640 * 480 * 3];
int tmp_buf_sz = 0;

VideoCapture cap;

#define LIBDLL extern "C" __declspec(dllexport)

LIBDLL int returnQ(float *data_img, int x)
{
    if(x==1)
    {
        cap.open(0);
        cap.set(CV_CAP_PROP_FRAME_WIDTH, 640);
        cap.set(CV_CAP_PROP_FRAME_HEIGHT, 480);
        cap.set(CAP_PROP_FPS, 30);
    }

    cap >> frameRGB;
    img = frameRGB.clone();

    split(img, ch);
    ch[0].convertTo(datas, CV_32F);
    datas = datas.reshape(1, 1).t();
    ch[1].convertTo(datas2, CV_32F);
    datas2 = datas2.reshape(1, 1).t();
    ch[2].convertTo(datas3, CV_32F);
    datas3 = datas3.reshape(1, 1).t();

    vconcat(datas, datas2, datas4);
    vconcat(datas4, datas3, datas4);

    tmp_buf_sz = 640 * 480 * 3;
    memcpy(data_img, datas4.data, sizeof(float)*tmp_buf_sz);

    waitKey(1);

    return *data_img;
}

And my code for Python:

import cv2
import ctypes
from ctypes import * 
import numpy as np

dll = cdll.LoadLibrary(r"./UVCStreamerTest.dll")
dll.returnQ.restype = c_int
data_img = (c_float * 921600)()

x=1

while(True):

    ss = dll.returnQ(data_img, x)
    if ss!=0:   

       x=0

       im = np.array(data_img)
       im = np.uint8(im)

       # Convert to BGR

        B = im[0:640*480]
        G = im[(640*480):(640*480*2)]
        R = im[(640*480*2):]

        B = B.reshape(480, 640)
        G = G.reshape(480, 640)
        R = R.reshape(480, 640)

        img = cv2.merge((B,G,R))    

        cv2.imshow("Image",img)
        cv2.waitKey(1)

Is there a way for better communication between C++ and Python?

When C++ has the task of activating the camera and sending images to Python.

This webcam call from C++ is just an example for the communication between C++ and Python. If it successful, I will change my camera which use specific library for UVC camera in C++. The flowchart will be the same, C++ capture image from camera and send to Python, then Python show the images.

Thank you in advance.

Edit: I found the solution. Only need to add: cap.open(0)

I edit my code above in case someone has the same problem.

Answer 1

Check out Boost.Python

I think it is more convient than the "cdll.LoadLibrary()"