APPLICATION NOTE
PCN-1001-00 Accuracy Rev 1.0 - July 2009 - ETH_PCN-1001-00 Accuracy_USM1.0
DIGITAL TECHNOLOGIES FOR A BETTER WORLD
www.eurotech.com
PCN-1001-00 Accuracy - Application Note
Disclaimer The information in this manual has been carefully checked and is believed to be accurate. Eurotech assumes no responsibility for any infringements on patents or other third party rights, which may result from its use. Eurotech assumes no responsibility for any inaccuracies that may be contained within this document, and makes no commitment to update or keep current the information contained within this manual. Eurotech reserves the right to make improvements to this document and/or product at any time, without prior notice.
Trademarks All trademarks - both marked and unmarked - appearing in this document are the property of their respective owners.
Technical Assistance If you have any technical questions or cannot isolate a problem with your device, please e-mail the Eurotech Technical Support Team:
[email protected].
© 2009 Eurotech Spa Eurotech S.p.A. A member of the Eurotech Group Via Fratelli Solari, 3/a 33020 - AMARO (UD) ITALY
2
PCN-1001-00 Accuracy - Application Note
Table of Contents PCN-1001-00 Accuracy .............................................................................................................................. 1 Disclaimer .................................................................................................................................................. 2 Trademarks ................................................................................................................................................ 2 Technical Assistance................................................................................................................................. 2 Table of Contents....................................................................................................................................... 3 Product overview ....................................................................................................................................... 4 PCN-1001-00 error definition ..................................................................................................................... 6 PCN-1001-00 characterization ................................................................................................................... 8 Error calculation: an example ..................................................................................................................10 Eurotech Group Worldwide presence......................................................................................................11
3
PCN-1001-00 Accuracy - Application Note
Product overview The PCN-1001 Passenger Counter is a compact and fully integrated device based on non-contact stereoscopic vision technology. It is able to count passengers entering or leaving busses, trams or trains with high accuracy. It also has the capability to record time and date information, which allows users to make a variety of statistical analyses at a later time.
PCN-1001-00
The device is specifically designed to be used in mobile environments. Thanks to its shape with a front mask that allows it to be mounted in an embedded position under doorways, it becomes almost invisible. The optical panel can also be angled to allow for mounting in locations that are uneven or not directly above the intended detection area. The Passenger Counter is an autonomous device with two stereoscopic cameras, which capture images of the area below the device. Using this information it then analyzes the height, shape and direction of any object within the detection area. The PCN-1001 comes with four integrated high luminosity infrared LEDs, which enable the system to operate in any kind of lighting condition, even in complete darkness. Thanks to the circular automotive connectors, it is possible to network PCN-1001 units together; this allows users to monitor all the doors within a vehicle in real time. If the width of a doorway is greater than 120 cm, two or more counters can be interfaced together. In this configuration, the first unit recognizes the presence of the others and manages any overlapping areas within the effective detection zone, which ensures that passenger movements are correctly analyzed. In all configurations, an RS-485 serial communications bus is used to interface the PCN-1001 units to the host PC of the vehicle. Isolated digital inputs and outputs can be used to control devices within the vehicle, which helps to guarantee its correct operation at all times. For example, the unit stops counting when the doors of the vehicle are closed. The Passenger Counter can be interfaced with other Eurotech mobile computing systems to create a complete intelligent and fully customizable network with advanced surveillance, communication and infotainment capabilities.
4
PCN-1001-00 Accuracy - Application Note
The system provides high accuracy due to the innovative technologies used; in real life operating conditions an accuracy of 97% is achievable. In the following paragraphs we are going to define how to measure and define errors in a statistical way as well as provide practical examples.
5
PCN-1001-00 Accuracy - Application Note
PCN-1001-00 error definition The counting accuracy of an Automatic Passenger Counting (APC) system is a fundamental consideration when choosing a device. It is well known that current people/passenger counters available on the market today indicate their error statistics using various methods that are not always clear; Eurotech, with this document, establishes precisely how the counting accuracy of the PCN-1001-00 is determined. The following represents the types of error as defined by Eurotech: • • •
Balanced counting error on entering people Balanced counting error on exiting people Unbalanced people counting error
•
Global people counting error
In order to define system errors, it is necessary to first establish that the PCN-1001-00 system is able to detect and count the traffic of people as defined as movement through its field of vision in a single direction which can be set-up as IN or OUT. If we indicate with
∑ In _ pcn
the number of registered passages by the PCN-1001-00 in the direction
people
configured as IN and
∑ In _ real and the number of actual passages in the direction configured as IN, we people
can define the Balanced counting error on entering people as:
∑ Balanced counting error on entering people =
In _ pcn −
people
∑ In _ real people
∑ In _ real people
If we indicate with
∑ Out _ pcn
the number of registered passages by the PCN-1001-00 in the direction
people
configured as OUT and
∑ Out _ real the number of actual passages in the direction configured as OUT, people
we can define the Balanced counting error on exiting people as:
∑ Out _ pcn − ∑ Out _ real Balanced counting error on exiting people =
people
people
∑ Out _ real people
6
PCN-1001-00 Accuracy - Application Note
In cases where the use of the PCN-1001-00 is in correlation to predetermined entrances and exits, for example: counting people in elevators or vehicles where the number of entrances must be equal to the number of exits, the total number of transiting passengers in both entrance and exit could be significant to pinpoint the concept that entering people must also exit. In this case we can define the Number of people as:
In _ pcn +
∑ Number of PCN counted people =
∑ Out _ pcn
people
people
= N_PCN
2
and
In _ real +
∑ Number of real counted people =
people
∑ Out _ real people
= N_real
2
In these cases the Global people counting error formula is preferred:
Global people counting error =
N _ PCN − N _ real N _ real
Finally we define the Unbalanced people counting error as:
∑ ( In _ PCN − In _ real ) + ∑ (Out _ PCN − Out _ real ) Unbalanced people counting error=
people
people
∑ ( In _ real + Out _ real ) people
When for the same application and regardless of the type of environment, closed or open space, more than one PCN-1001-00 is used, it is necessary to expand the sum of the previous formula considering the counts included by all PCN-1001-00s. Example: evaluation of balanced counting error on people entering a 2 door bus, based on a test run of 10 stops. 10
2
∑ ∑ ∑
Balanced counting error on entering people =
In _ pcn −
stop =1 door =1 people 10
10
2
∑ ∑ ∑ In _ real
stop =1 door =1 people 2
∑ ∑ ∑ In _ real
stop =1 door =1 people
7
PCN-1001-00 Accuracy - Application Note
PCN-1001-00 characterization Eurotech applied these formulas and the concepts described above using a PCN-1001-00 in a test laboratory. The tests were carried out in a closed environment, with the PCN-1001-00 installed above an artificial doorway in normal, internal lighting conditions. Fraudulent crossings, intended to provoke counting errors were not simulated. For there tests the following types of crossing have been considered: • • • • •
Single crossing Double crossing, same direction Double crossing, different direction Triple crossing, same direction Triple crossing, different direction (2 vs. 1)
Within this type of crossing the following distances, from person to sensor, have been allowed: • • • • • • • • •
20cm 30cm 45cm 60cm 75cm 90cm 105cm 120cm 135cm
For non-single crossings people of uniform height were used. For triple crossings, all three people were of the same height. The tests were carried out using two doorway widths, 120cm and 140cm. For each type of crossing, each distance from head-sensor and each passage width, the test was carried out with “Distance Configuration” set as “25-30 cm” and “31-40 cm”. For each specific configuration and type of test at least 50 crossings in each direction were performed. The final results were constructed considering a uniform distribution regarding the 5 types of crossings. From the tests carried out it can be said that the error tendency is towards missed counts and not false counts. In the following graphs it is possible to track performance related to the head-sensor distance. The graphs summarize the results both by taking into account the error balancing and by not taking into account the error balancing. The results reflect a Distance Configuration of “25-30 cm” and “31-40 cm”. Each graph shows the accuracy with an opening width of 120cm and 140cm.
8
PCN-1001-00 Accuracy - Application Note
Performance %
Unbalanced Performance (Distance Configuration 25-30) 96,6 94,4
93,8
95
99,8 99,7
99,6 98,8
99 97,8
96
93,6
93,6
91,2
85,6
85 75
78,6 73,2
73,8
65 57,6 55 0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150 Gate 120cm
Distance People Head - Sens or (cm )
Gate 140cm
Performance %
Balanced Performance (Distance Configuration 25-30) 98,2
95,8
95
95,2
99,8 98,8
99,8 98,4
99,4 98,2
98,4 94
85,6
85
94,4 93,2 80,2 73,6
75,6
75 65
62,4
55 0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150 Gate 120cm
Distance Pe ople Head - Sensor (cm )
Gate 140cm
Performance %
Unbalanced Performance (Distance Configuration 31-40)
95,0 85,0
99,8 99,7
99,6 98,8
96,6 94,4
91,3
99 97,8
96 93,6
94,5 91,4 87,4
91,9
82,1
75,0 67,2
65,0 55,0 0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150 Gate 120cm
Dis tance People He ad - Sensor (cm )
Gate 140cm
Balanced Performance (Distance Configuration 31-40)
Performance %
98,2 95,0
92,4
85,0
95,2
99,8 98,8
99,8 98,4
99,4 98,2
98,4 94
97,1 93,8
93,6 88,7
83,2
75,0 70,4 65,0 58,3
55,0 0
10
20
30
40
50
60
70
80
90
100
110
120
Distance People Head - Se nsor (cm )
130
140
150
Gate 120cm Gate 140cm
Figure 1 Passage accuracy PCN-1001-00 For applications in different testing environments, it would be necessary to re-evaluate the related variables that could influence statistics.
9
PCN-1001-00 Accuracy - Application Note
Error calculation: an example The following table shows the entrance and exit data of a possible route with 10 stops. Real counting
PCN-1001-00 counting
Stops 1
In
Out
In
Out
27
0
24
0
2
9
3
9
4
3
16
15
18
14
4
11
9
11
9
5
9
18
9
19
6
12
14
10
14
7
4
7
4
7
8
8
15
8
15
9
4
12
4
12
10
0
7
0
7
total
100
100
97
101
Below we calculate the errors referring to the data in the above table:
Global counting error =
(97 + 101) − (100 + 100) x 100% = - 1% 100 + 100
Balanced counting error on people entering =
Balanced counting error on people exiting =
97 − 100 x 100% = - 3% 100
101 − 100 x 100% = 1% 100
In some cases manual counting cannot be performed, therefore the error calculation method is the average deviation. This is defined as the relation between the difference of the absolute value of the entrance and exit values as counted by the PCN-1001-00. Using the example data we find:
Mean deviation o=
10
97 − 101 97 + 101
x 100% = 2%
Eurotech Group Worldwide presence
Americas
Europe
Asia
North America
Central & Southern Europe
Japan
EUROTECH Inc toll free +1 888.941.2224 tel. +1 301.490.4007 fax +1 301.490.4582 e-mail:
[email protected] e-mail:
[email protected] web: www.eurotech-inc.com
EUROTECH Italy tel. +39 0433.485.411 fax +39 0433.485.499 e-mail:
[email protected] e-mail:
[email protected] web: www.eurotech.com
ADVANET tel. +81 86.245.2861 fax +81 86.245.2860 e-mail:
[email protected] web: www.advanet.co.jp
Western Europe
China
EUROTECH UK tel. +44 (0) 1223.403410 fax +44 (0) 1223.410457 e-mail:
[email protected] e-mail:
[email protected] web: www.eurotech.com
VANTRON tel. + 86 28.85.12.39.30 fax +86 28.85.12.39.35 e-mail:
[email protected] e-mail:
[email protected] web: www.vantrontech.com.cn
PARVUS CORPORATION tel. +1 800.483.3152 fax +1 801.483.1523 e-mail:
[email protected] e-mail:
[email protected] web: www.parvus.com
EUROTECH France tel. +33 04.72.89.00.90 fax +33 04.78.70.08.24 e-mail:
[email protected] e-mail:
[email protected] web: www.eurotech.com Northern & Eastern Europe EUROTECH Finland tel. +358 9.477.888.0 fax +358 9.477.888.99 e-mail:
[email protected] e-mail:
[email protected] web: www.eurotech.com
www.eurotech.com