From the course: AR Development Techniques 03: Recognition and Tracking

How does it work?

From the course: AR Development Techniques 03: Recognition and Tracking

Start my 1-month free trial

How does it work?

“

- [Teacher] In this video, I'm going to answer three questions. What does image tracking? How does it work? And what kind of images are suitable for tracking? Image tracking is the process of recognition of an image and tracking that image for each camera frame. A virtual object, like a virtual 3D model, can be overlaid on that image so that the model moves along with it. To create an app with image tracking functionality, we feed the image, which has to be tracked, to our project beforehand. Behind the scenes, the algorithms inside the IR package recognize unique patterns or feature points on that image based on the texture of that image. So for example, if we feed this image to the image recognition algorithm, it will try to find a pattern or unique feature points on this image. Once the project is compiled and the app is deployed on our device, then whenever the app sees an image, it will try to compare the pattern with the image which already exists in the project database. If the pattern is same, then the image gets recognized and it can be tracked. Since the algorithm needs to recognize unique patterns and feature points, the image that you should use should be as textured or as complex as possible. Have a look at these three images. The first one is the best one because it is quite textured, and so it has a lot of feature points. The second one is not that great because it will have very less patterns or feature points, and so the tracking will always be unstable. The third one is not usable at all because there are no patterns to be recognized. And it's not because the image is white. The color of the image does not matter. It can be any other color and still it will have no feature points. So that's all for this video. In the next video, we'll create a simple image tracking app.

Contents