How to correctly grab matrix from Queue (opencv) - opencv

Question

I try to make frame grabber class from IP Camera using std::queue and opencv
In this class, It takes and store maximum 10 frame,
and Other class, get frame from buffer.
/* Camera.cpp */
struct FRAME
{
cv::Mat Img;
uint64_t Time;
bool ReadError;
};
class Camera
{
std::queue<FRAME> m_buffer;
std::mutex m_buffer_mutex;
...
};
void Camera::threadFunc_get_stream_from_camera()
{
while (true)
{
FRAME frame;
... // read sequence (if cv::read is fail, store frame routine is skipping)
m_buffer_mutex.lock();
while (m_buffer.size() >= 10)
m_buffer.pop();
m_buffer.push(frame);
m_buffer_mutex.unlock();
}
}
FRAME Camera::grab()
{
std::unique_lock ulock(m_buffer_mutex);
FRAME frame;
if (m_buffer.empty() == false)
{
frame = m_buffer.front();
m_buffer.pop();
return frame;
}
}
/***************/
/* Testing.cpp */
int main()
{
// Camera* cam ..
...
while (true)
{
auto frame = cam->grab();
if (frame.Img.data == nullptr || frame.Time == 0)
continue;
... // frame image processing
}
}
I kept the program this way for about 3 days, and no special bugs or memory sticks were found.
But the reason I'm asking this question is,
whether it's right to use non-pointer(FRAME) in grab() function.
I'm using C++17, and I think there is no problem because the uint64_t or bool values in the structure are copied in this version.
but, How about cv::Mat ?
I don't want to using Producer-Consumer pattern, because not want to contain testing class side's code in Camera class.