Log in or register to post comments

Increasing detection distance

March 6, 2020 - 3:58am #1

We are evaluating Vuforia for an AR project that needs to track around 10 markers simultaneously. We have been impressed with Vuforias tracking abilities, but the initial detection of a VuMark appears to require the VuMark to be very close to the camera. For the user this means they have to lean close in to each marker for the initial detection, and then the tracking works pretty well after that. Is the initial detection distance configurable in any way? What we suspect is happening is that Vuforia is doing an analysis of the camera image to find candidate contours to check for Vumark codes. Any contour below a certain size is probably rejected, and this is why you need to be close for the initial detection. Is this the case, and is there any way to configure that? We understand that this would have a performance cost as there could be many more candidate contours in the image to check for Vumarks.

 

The performance we are looking for in terms of initial detection is similar to that of Aruco marker detection as seen here:

 

https://docs.opencv.org/3.4/d5/dae/tutorial_aruco_detection.html

 

Would it be possible to get similar performance from Vuforia if using a similar marker pattern? We have been testing using the Mars samples.

Increasing detection distance

March 10, 2020 - 7:01am #2

Hi,

The Vuforia SDK extracts feature points from the marker with each camera frame. Thus, the more distant the marker becomes from the camera, the more difficult it will be for the algorithm to provide an accurate pose. You can estimate 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 detectable up to a distance of about 2 meters (20 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, device camera properties, and target rating.

The Extended Tracking feature can be used to supplement our 2D tracker's performance and overcome some of these limitations. It also extracts features, from the environment in addition to the target, so may help to stabilize your experience.

So the detection range depends on the size of the VuMark. For ex. using the VuMarks from the Unity Core sample apps which is 12 x 7 cm, I was able to detect them from around 30 cm without any issues.

printout available here: https://library.vuforia.com/content/dam/vuforia-library/docs/target_samples/unity/mars_vumarks.pdf

Unity Core sample app is available through the Unity asset store

Thank you.

Vuforia Engine Support

 

Log in or register to post comments