Eye Tracking System Using Pattern Recognition Algorithm and LabView

As the technology advances, need for an interactive system has increased since then. A system which responds to bio signals transmitted by user is dream project of lot of scientist and engineers. Eye tracking is one of the essential components of that project. It is being employed in mobile phones and advanced computer for a plethora of applications.

Eye tracking requires vigorous image processing, as image constitutes various other variables apart from the eyes. All those variables are required to be filtered out and only the region of interest needs to be tracked down. Once a single image is filtered, a series of images are tested down for the same.

Our project has used these principles and implemented them using LabVIEW.

Initially user needs to define region of interest and later it tracks them in real time feed.

Here is the block diagrams, described individually there functions. We have used the vision tools available on LabView.

1.1 Continuous Acquisition

In this particular implementation, we have used Vision and motion basic palette functions like ‘Open camera’ and ‘Configure’.  

The output from the acquisition is divided in two parts. First to Snap, where the continuous stream is taken one frame at a time.

2.2 Snap (One frame at a time)

The Frame acquired is converted to grey scale image using ‘Cast Image’ block as shown in the figure. This image is saved in a local variable. Snap is taken only when the ‘Snap’ button is pressed by the user.

The image is saved in local variable.

2.3 Get ROI

The user then pressed Get ROI button, where separate widow pops up. Using the function ‘Select rectangle’, the eyes from the image is extracted and given to ROI descriptor. The ‘learn pattern’ function recognizes the ROI and saves it in a template.

2.4 Tracking

In this structure, the images are acquired directly from webcam, converted in grey scale using ‘Cast Image’. It is then carry forwarded to ‘IMAQ Setup Match Pattern’ where the template is matched with coming online feed, using ‘IMAQ match pattern 2’. Then we extract bounding box values and given to ‘IMAQ Overlay Rectangle’. This block will overlap the bounding box values on the online feed.  

The output generated is as follows:

Figure 1: The output of the eye tracking system.

The code will be available to blog followers upon request.