1.0 Introduction
Base case: staffs pass through one at a time in one line at varying walking speed
-- success rate to exclude staff is >80%, regardless of frame rate fluctuation (since system grab frames one by one from buffer & Tof readings is retrieved during each processing cycle)
Case 1: Two or more staffs walk through side by side
Exclusion success 80% of the time as long as the staffs do pass through the ToF coverage zone (mentioned in Device Installation Guide)
Case 2: Staff and non-staff walk through in opposite direction
Exclusion success rate is similar to base case if they intersect NOT within the ToF coverage zone. Mismatch will happen if they intersect or stop within ToF zone (either staff not excluded but non-staff excluded or both of them get excluded). Other than that, if staffs walk through directly under the counter for them to be excluded, people coming in the opposite direction would naturally stop and wait or walk sideways to avoid collision, hence no longer intersect within Tof zone.
1.1 Limitations
-
Exclusion of NON-staff might happen if the separation distance between staff and non-staff when they pass through one after another is <1m. Adjusting the matching distance & threshold can help to lower the chances of mismatch, but with trade-off to staff exlusion success rate. Possible to improve further in future by doing timestamp matching between frames captured and ToF readings.
-
Still not a good fit to exclude staff staying directly under the counter for extended period of time given the current ToF sensor resolution (8x8). The staff (with the reflective tag on) movement can cause passerby to be excluded as well, especially when staff is occupying part of the tof sensor coverage zone and customers pass by within the zone. (Higher ToF sensor resolution and wider coverage would make this feasible)
1.2 Other potential use cases of ToF sensor on 5D Pro
-
For noise reduction based on ToF sensor distance value readings and comparison with minimum person height
-
For height estimation
-
Alternative for narrow entrance counting in low light environment