Skip to main content

Tracking Config

MindAR comes with the following configurable parameters to fine tune the tracking.

Smoothing Control​

Jittering is a common issue in AR applications. The library computes the position of the targets in every frames. The problem is that there are always slight variations across aframes (e.g. noises coming from camera input or computational errors) even though you may think you are holding the devices steadily. Result is that the augmented contents will appear shaky.

One solution to smooth the jittering is to apply filtering. At a high level, it means taking some kind of rolling average of multiple frames to position the targets. The contents will then appeared more stable.

However, smoothing also comes with a price. Since we are using the previous N frames to interpolate the position, there will be a delay for the content to move to the latest position. This is more obvious if you move the camera quickly.

MindAR implements OneEuroFilter. There are two adjustable parameters called cutoff frequency (filterMinCF) and speed coefficient (filterBeta). In general, decreasing the value of filterMinCF can reduce the jittering and increasing the value of filterBeta reduce the delay. They are, however, somehow fighting against each others.

They default values of filterMinCF and filterBeta are 0.001 and 1000. You can change them by specifying these parameters in mindar-image attribute. e.g.

<a-scene mindar-image="filterMinCF:0.1; filterBeta: 10"/>

WarmUp Tolerance​

By default, there is a small intentional delay to trigger the target found event to avoid false positive. More specifically, it requires the target image being detected in a continuos of warmupTolerance frames to be considered a success. The default value of warmupTolerance is 5, and you can change it by specifying this parameter in the mindar-image attribute, e.g.

<a-scene mindar-image="warmupTolerance: 1"/>

Miss Tolerance​

Similar, there is also a small intentional delay to trigger the target lost event. It requires the target image being un-detected in a continuous of missTolerance frames. The default value of missTolerance is 5, and you can change it by specifying this parameter in the mindar-image attribute, e.g.

<a-scene mindar-image="missTolerance: 1"/>