Noise

noise is a result of issues with the tracking quality from a troubleshooting perspective this means that the following calibrations aren't relevent in trying to resolve it cambar offset lens calibration delay noise is a result of either poor tracking environment substandard cambar calibration the steps for addressing troubleshooting differ slightly between natural and marker modes but in both cases we need to essentially understand whether noise is being generated by the cambar or the environment natural tracking test observe tracking features as a first step it's important to consider the current tracking environment the cambar is tracking in is there visible movement in the tracking features detected in the ui can the cambar be orientated to find more stable tracking features can the additional features be added to the environment to support tracking test pointcloud cleanup does the tracking quality significantly improve after a "pointcloud cleanup" is conducted? this can indicate that the pointcloud was containing too many incorrect tracking features (lighting changes, moving objects etc) for longterm tracking consider locking the pointcloud or moving to marker mode test check for unexpected loss of tracking features observe the detected tracking features in the ui, are stable tracking features dropping in and out of detection? this is a strong indicator that there is an issue with the cambar calibration test check pixel error the pixel error can provide an indication of the cambar calibration, when starting the tracking, is the initial value average or higher? this can indicate that the cambar calibration is poor the pixel error alone should not be used as the single indicator of cambar quality however, as the environment and type of features being tracked can also impact pixel error marker when operating in marker mode, as the system is using fixed markers in a locked pointcloud, it's less likely that noise is being generated from pointcloud pollution test check for movement in detected markers the marker tracking core looks for the centre of markers, check to see if any visible "wobble" or movement can be seen in the ui adjust cambar lighting to see if detection of the marker can be stabilised consider physically removing specific markers test check for intermittantly detected markers are some markers dropping in and out of detection when the camera is stationairy? this can be typcially caused by insufficent ir lighting or falloff towards the edge of the camera frame markers that drop in and out of detection, particularly when only tracking a handful of markers can introduce noise can cambar lighting be adjusted to stabilise marker detection? can larger markers be used to reduce ir light falloff? test cover one lens does noise reduce significantly when covering one of the cambar cameras? this is a strong indication that the cambar calibration is poor test is noise present everywhere? is noise visible in the camera tracking across the whole marker area? if it's only present in certain parts of the environment then it is likely caused by poor marker placement/detection instead of cambar calibration in the areas where noise is present is there a good (even) distribution of markers across the cambar view? a cluster of markers only in the corner or part of the frame will be more likely to cause noise are some markers intermittently detected markers that are detected poorly will introduce noise consider relearning the marker map or if required, removing troublesome markers physically from the environment are there "false" markers detected can you observe falsely detected markers? these could be reflections or lights in the environment look to see how these can be excluded from the tracking map