Threading.Timer stack overflow - wpf

Question

Hy, I just wanted to find out how to implement a loop that takes around 80 to 120ms (mostly 80ms) to execute. The loop has to execute for about 30 minutes... basically it is a SURF matching algorithm.
Currently I am using a System.Threading.Timer to create a timer that executes after every 90ms, but the problem is that since the computation time is variable, so after some time the stack overflows and the program closes.
I'm using WPF to create the GUI.
Is there a better way to implement such a loop using threading? Any help would be much appreciated! Thanks in advance.
//initialization
private System.Threading.Timer Visual_AR1;
Visual_AR1 = new System.Threading.Timer(new TimerCallback(Video_AR1), null, 0, 90);
private void Video_AR1(object state)
{
lock (this)
{
// SURF matching
modelImage_AR1 = new Image<Gray, byte>(droneControl_AR1.BitmapImage).Resize(1.8, INTER.CV_INTER_NN);
map_image_d1 = DrawMatches_gpu.GPU_0(wayx, modelImage_AR1, observedImage, pgpuObservedKeyPoints_imp, pgpuObservedDescriptors_imp, out matchTime_0, out pX1, out pY1);
Dispatcher.BeginInvoke((Action)delegate()
{
label4.Content = "Time Taken by GPU_0 : " + matchTime_0.ToString();
});
mask_selector_d1();
}
}
is this a viable solution ?
private Thread threadTask = null;
private void threadTask_Start()
{
if (threadTask == null) {
threadTask = new Thread(SURF);
threadTask.Start();
}
}
private void SURF()
{
while(true)
{
lock (this)
{
// SURF matching
modelImage_AR1 = new Image<Gray, byte>(droneControl_AR1.BitmapImage).Resize(1.8, INTER.CV_INTER_NN);
map_image_d1 = DrawMatches_gpu.GPU_0(wayx, modelImage_AR1, observedImage, pgpuObservedKeyPoints_imp, pgpuObservedDescriptors_imp, out matchTime_0, out pX1, out pY1);
Dispatcher.BeginInvoke((Action)delegate()
{
label4.Content = "Time Taken by GPU_0 : " + matchTime_0.ToString();
});
mask_selector_d1();
}
thread.sleep(40);
}
}

Answer 1

Instead of making the timer tick perform the work, make the timer tick queue up a job and have something else (presumably threads from the thread pool) consume the jobs. You can use a semaphore to throttle the number of jobs running simultaneously and you can examine the state of the queue to avoid this sort of over-commit problem.