![background image](http://html1.mh-extra.com/html/sick/1088784/1088784_operating-instructions-manual_1264409170.webp)
9
ANNEX
170
8020776/12AX / 2019-05-31|SICK
O P E R A T I N G I N S T R U C T I O N S | Free Flow Profiler
Subject to change without notice
9.4
Terminology relating to statistical measurement accuracy
The statistical measurement accuracy of the system involves two different quantities: the
accuracy of the counting process and the accuracy of the classification process. As these
operating instructions repeatedly refer to these quantities, a simple example will now be
used to demonstrate the difference between the two.
In the example, measurement accuracy is determined on the basis of 10,000 vehicles.
For ease of understanding, the vehicles consist of cars (5,000) and trucks (5,000) only.
Fig. 76: Counting accuracy and classification accuracy
The counting accuracy shows how many of the vehicles have been correctly detected and
– consequently – counted. Counting accuracy is independent of the type of vehicles in the
traffic.
In our example, 20 of the 10,000 vehicles were not detected. This gives a counting
accuracy of 99.8%.
The classification accuracy shows how many of the vehicles have been correctly
classified. In our example, 9,800 of the 10,000 vehicles have been correctly classified.
The system’s classification accuracy amounts to 98%.
The classification rate for trucks is higher than for cars.
•
Trucks: 4,910 of 5,000 vehicles were correctly classified = 98.2%
•
Cars: 4,890 of 5,000 vehicles were correctly classified = 97.8%
The classification accuracy of the entire system therefore depends on the type of vehicles
in the traffic and the category system used. Depending on the type of vehicles in the
traffic and the category system used, the measurement accuracy of the entire system may
increase or decrease.
For most of the data recorded for a vehicle, there is a confidence level (between 0 and
100%). The confidence level expresses the level of certainty with which the value was
correctly identified by the system component.
Fig. 77: Output data with confidence level
The confidence level can be processed internally as well as during analysis in downstream
systems.
Counting accuracy
Classification
accuracy
Confidence