Accuracy in outdoor augmented reality is a budget decision, and this study lays out the numbers.
A seven metre error is the difference between a marker sitting on a shop door and a marker floating over the middle of the road. That is roughly the gap between a conventional outdoor AR setup and the combined VSLAM and Street View approach tested here.
Outdoor AR has quietly become ordinary. Campus tours, transit stops, heritage trails and retail promotions all lean on it. When the anchor is wrong, users do not read it as a sensor limit. They read it as a broken app.
Sensor fusion gives an absolute position, but a soft one: the GPS fix wobbles and the compass follows magnetic noise. Visual SLAM is excellent at relative motion and poor at holding an absolute frame, because small errors accumulate as the user walks. Street View matching is the opposite: slow and sparse, but it lands on a fixed coordinate.
Used together, each layer covers the weakness of the others. That lesson generalises beyond AR. If two weak signals fail in different ways, combining them costs less than making either one strong.
For a phone two or three years old, doubling the memory footprint of a camera screen is a real constraint. It shows up as heat, battery drain and dropped frames. The authors are direct: accuracy costs load, and the right choice depends on the app.
Check Street View coverage for your target streets before designing around it. Keep the sensor fusion path as a fallback. And decide early whether you need a marker on a doorway or just a marker in the right block, because that sets the resource budget.
Brata et al. (2024). An enhancement of outdoor location-based augmented reality anchor precision through VSLAM and Google Street View. Sensors, 24(4), 1161.
