How can I improve detection and tracking stability

March 24, 2014 - 3:09am #1

If you are experiencing detection or tracking performance issues, such as difficult detection of a target, target getting lost eaily, or unstable / jittery tracking, you should check the following elements:

Target "star rating": 

Image Targets are detected based on natural features that are analyzed in the target image, stored in a database and then compared at run-time with features in the live camera image. The star rating of a target ranges between 1 and 5 stars; although targets with low rating (e.g. 1 or 2 stars) can usually detect and track well, for best results you should try and aim for targets with 4 or 5 stars. To create a trackable that is accurately detected use images that are:

  • Rich in detail (Ex. street-scene, group of people, collages and mixtures of items, sport scenes),
  • Has good contrast i.e. it has both bright and dark regions, is well lit and not dull in brightness or color
  • Does not have repetitive patterns such as a grassy field, the front of a modern house with identical windows, and other regular grids and patterns.

For more information on Image Targets features and rating and for advice on how to enhance your Image Targets, please consult these pages:

https://developer.vuforia.com/resources/dev-guide/natural-features-and-rating

https://developer.vuforia.com/resources/dev-guide/image-target-enhancement-tricks

 

Camera Focus Modes:

If the target is not well in-focus in the camera view, the camera image might result blurry and the target details may be hard to detect; as a consequence, detection and tracking performance may be negatively affected.

 It is therefore recommended to make use of the appropriate Camera Focus Mode to ensure the best camera focus conditions; in particular, the continuous autofocus mode (FOCUS_MODE_CONTINUOUS_AUTO) is the recommended option, as it allows your device to automatically adjust the focus as the view changes; however, not all devices support this mode, so you may also want to consider the other focus modes available in the Vuforia API; for a complete description of camera focus modes, please see this article:

https://developer.vuforia.com/resources/dev-guide/continuous-autofocus-and-other-focus-modes

 

Lighting conditions:

The lighting conditions in your test environment can significantly affect target detection and tracking; make sure that there is enough light in your room or operating environment, so that the scene details and target features are well visible in the camera view.

Also, consider that Vuforia is generally working best in indoor environments.

 

Target size:

For tabletop, near-field, product shelf and similar scenarios, a physical printed image target should be at least 5 inches or 12 cm in width and of reasonable height for a good AR experience. The recommended size varies based on the actual target rating and the distance to the physical image target. Consider increasing the size of your targets if the distance of the target is higher. As a very rough "rule-of-thumb", you can get an idea of the minimum size that your target should have by dividing your camera-to-target distance by ~10. For instance, a 20 cm wide target would be typically detectable up to a distance of about 2 meters (2 cm x 10); note, however, that this is just a rough indication and the actual working distance/size ratio can vary based on lighting conditions, camera focus and target rating.  

 

Viewing angle:

If you are looking at the target from a very steep angle (i.e. if your target appears very "oblique" w.r.t the camera), the target features will be harder to detect nd tracking may also be less stable. When defining your use scenarios, keep in mind that a target facing the camera (i.e. whose normal is well aligned with the camera viewing direction) will have better chances to get detected and tracked.

 

 

Topic locked