Setting roi to eyes and mouth only in camera capture and further processing

Question

I am making a drowsiness detection system. so for that i want to set the ROI of the camera capture to eyes and mouth only and i further want to note their characteristics. i am able to set the ROI manually but i want the ROI to be set automatically to the eyes only. further after setting the ROI i want to further process it by applying the hough transform.is there any way to do that?

Answer 1

In Opencv library you can find and use the cascade classifier to detect objects in video sequences. It is based on Haar-like features and boosting algorithm. Opencv gives you pre-learnt cascade to detect faces, eyes, mouth, nose. All these cascades are learnt on huge databases and are quite efficient for faces in near frontal view condition.

You should use cascades to detect facial features you want to analyse and then, look for a tracking procedure (such as meanshit or bayesian filtering) in order to be robust to different pose during the video sequence.

Opencv also implements the Hough transform. Here for opencv code.