Kinect DK obtains color RGB images in cv:: mat format (used in openpose)

The main question that this blog solves , obtain Kinect dk Medium color map , And put the color map into openpose in , Get the hand bone points . And then from kinect dk Get the depth value corresponding to each bone point in the depth map of , So as to get three-dimensional bone points . Here are the problems encountered , And corresponding solutions ：
problem 1：k4a::image turn CV::Mat
Use cv::Mat The parameterized constructor of

k4a::capture capture;
cv::Mat m_colorMat;
if (m_device.get_capture(&capture)) {
	k4a::image m_colorImage = capture.get_color_image();
	cv_rgbImage_with_alpha = cv::Mat(m_colorImage.get_height_pixels(), m_colorMat.get_width_pixels(), CV_8UC4,
						(void *)m_colorImage.get_buffer(), cv::Mat::AUTO_STEP); 
cv::imshow("color", m_colorMat);
}

problem 2：CV::Mat Medium RGBA turn RGB
step[0] Is the number of bytes in a row of elements in a matrix
step[1] Is the number of bytes of an element in the matrix
Openpose The format you want in is RGB（ The pictures obtained by the default camera can be recognized normally , Refer to its format ）

What you get directly is RGBA

operation ： transformation Mat Type of image

cv::cvtColor(m_colorMat, m_colorMat, cv::COLOR_BGRA2BGR); //  transformation 

problem 3： Picture input to openpose Cannot detect normally
The analysis reason Mat.data There is a problem with the data format in

k4a::image The data storage in is uint8_t,opencv It uses uchar, Use cv::imshow It can also be displayed normally （ problem 1 The code in will k4a::image Turn into cv::Mat after ）

openpose Data for cvInputData, requirement uchar format

Ideas ： First, put the data into uint_t To uchar, then k4a::image Turn into cv::Mat

//  Data transformation 
std::vector<uchar> pos; //  spot 
int width = m_colorImage.get_width_pixels();
int height = m_colorImage.get_height_pixels();
pos.resize(width * height * 4);
auto depth = m_colorImage.get_buffer();
memcpy(pos.data(), depth, width * height * sizeof(uchar) * 4);

//  obtain Mat
m_colorMat= cv::Mat(m_colorImage.get_height_pixels(), m_colorImage.get_width_pixels(), CV_8UC4,
					(void *)pos.data(), cv::Mat::AUTO_STEP); // m_colorImage.get_buffer()
// RGBA turn RGB
cv::cvtColor(m_colorMat, m_colorMat, cv::COLOR_BGRA2BGR);
std::move(pos);

result