Skip to main content

Table 3 Performance comparison of image-based localization approaches for the “patton” model

From: Fast and scalable structure-from-motion based localization for high-precision mobile augmented reality systems

Package HD4AR HD4AR with caching approach
Cache size - 1000 2000 5000 10,000
Localization 49/50 40/50 44/50 49/50 49/50
success-ratio (98 %) (80 %) (88 %) (98 %) (98 %)
Mean number of 2145 134 228 438 748
2D-to-3D matches      
Mean re-projection error 0.812 pixels 0.962 pixels 0.927 pixels 1.047 pixels 1.060 pixels
Mean localization time 2.312 sec 1.314 sec 1.484 sec 1.692 sec 1.836 sec
(sequential requests) (1 ×) (1.760 ×) (1.558 ×) (1.366 ×) (1.259 ×)
Mean localization time 0.754 sec 0.477 sec 0.547 sec 0.583 sec 0.627 sec
(parallel requests) a (1 ×) (1.581 ×) (1.378 ×) (1.378 ×) (1.203 ×)
  1. aHandled 4 concurrent requests by multi-threading