Detekce směru pohledu řidiče v obrazech

Abstract

This thesis deals with the detection of the view direction. The methods for monitoring determining the direction of view are includes in the thesis. Next, there are described methods of image processing, the problems associated with the detection of the eye and the direction of view. For the actual implementation is selected the method uses one IR camera. The head is detected using cascade classifiers and landmarks. Detection centre of the eye (pupil) is performed using an algorithm based on the gradient direction. The calculation of the view direction is calculated then as the deflection of the centre of pupil from the real centre of the eye. The algorithm has proved using test videos with an accuracy of up to 4◦. The created solution enables to detect the direction of view from the IR camera and tune the detection accuracy using the configuration file. The benefit of this work is that we are able to detect the direction of the driver’s view only from the camera.

Description

Import 23/08/2017

Subject(s)

detection, direction of view, OpenCV, LBP, HOG, HAAR, landmark, gradient orientation, detector

Citation