A55 North Wales Tunnels Area Video Automatic Incident Detection

advertisement
EasyWay
Evaluation Expert Group (EW EEG) Document – EEG/10/7
A55 North Wales Tunnels Area
Video Automatic Incident
Detection
For Use by Project Partners
Issue: 1st edition
Version: 0.1
Date: June 2011
EasyWay Evaluation of Video Automatic
Incident Detection
-i-
June 2011
Document details:
Date of preparation:
June 2011
Abstract:
Document number:
5083687/493/03/35695
Document version:
Issue 1
Status:
Issued for general use within EasyWay projects
Dissemination level:
Public
Distribution:
EasyWay projects
Key audience:
Project evaluation managers
European Commission
Associated texts:
Name
Authors:
Stuart Hancocks
Organisation/ Telephone
Project
Traffic Wales +44 1492 564712
Stephen Evans
Traffic Wales
+44 1492 564727
Email
Stuart.hancocks@trafficwales.com
Stephen.evans@trafficwales.com
Document Control Record:
Version
Main Changes:
Issue 1
Initial issue for comment
EasyWay Evaluation of Video Automatic
Incident Detection
Contributions from:
Date:
June 2011
- ii -
June 2011
PREFACE
This reporting template is designed to help those evaluating projects in EasyWay to prepare and
present the results of their work on evaluating the implementation of ITS projects in a common format.
It is important to recognise that projects in EasyWay are funded by Member States as well as by the
European Commission, and that some Member States will have their own requirements for presenting
evaluation results. However, adoption of this common format for presenting results will help to ensure
that results of similar implementations can be compared, assisting in the transfer of results between
sites and providing European Added Value. It will also ensure that a periodic report can be prepared
by the EasyWay Evaluation Expert Group (EW EEG) of progress towards achieving EasyWay’s
objective of reducing congestion and emissions from transport and improving safety.
Because the final peer reviewed version of the results will be made public, reports should not contain
any information which cannot be made available in the public domain.
EasyWay Evaluation of Video Automatic
Incident Detection
- iii -
June 2011
TABLE OF CONTENTS
Document Control Record:
ii
REPORT TEMPLATE
1
Project Reference (From EasyWay Work plan):
1
Project Name: A55 Tunnels Area Video Incident Detection Upgrade
1
EasyWay Region: StreetWise – Wales, UK
1
1. Area For Review
2
1.1.
Site
2
1.2.
Issues Addressed
2
2. Description of the ITS Project
4
2.1.
Project Domain
4
2.2.
Objectives
4
2.3.
Systems and Technologies Applied
5
2.4.
Status of the Project
8
3. Key Evaluation Results
9
3.1.
Impact on Traffic Flow
9
3.2.
Impact on Safety
10
3.3.
Impact on Environment
10
3.4.
Other Key Results
10
4. Evaluation Planned
11
4.1.
Timing and Type of Evaluation
11
4.2.
Objectives for the Evaluation
11
4.3.
Research Questions
11
EasyWay Evaluation of Video Automatic
Incident Detection
- iv -
June 2011
4.4.
Study Area for the Evaluation
11
4.5.
Expected Impacts
12
5. The Impact of the Project - Results
13
5.1.
Technical Performance
13
5.2.
Results
13
5.3.
Reliability of Results
16
5.4.
Research Questions Answered
16
5.4.1.
EFFICIENCY
16
5.4.2.
EFFECTIVENESS
16
5.4.3.
OPERATOR USABILITY AND PERCEPTION
17
5.4.4.
FLEXIBILITY
17
5.4.5.
VALUE FOR MONEY
18
5.4.6.
POSSIBLE IMPROVEMENTS
18
5.5.
5.6.
Overall Assessment
18
5.5.1.
SAFETY
18
5.5.2.
EFFICIENCY
18
5.5.3.
ENVIRONMENT
19
Additional Aspects
19
6. European Dimension: Transferability of the Results
20
Annex 1: Technical Annex
20
6.1.
Breakdown of Project Costs
20
6.2.
Data Collection Methods
21
EasyWay Evaluation of Video Automatic
Incident Detection
-v-
June 2011
REPORT TEMPLATE
Project Reference (From EasyWay Work plan):
Project Name: A55 Tunnels Area Video Incident Detection Upgrade
EasyWay Region: StreetWise – Wales, UK
EasyWay Evaluation of Video Automatic
Incident Detection
-1-
June 2011
1.
Area For Review
1.1. Site
The site is the tunnels area on the A55 between junctions 15 and 18 in North Wales, United
Kingdom. The A55 forms part of the Trans-European Transport Network (TEN-T), which links
Ireland with the rest of Europe. The A55 is also the main transportation artery for North Wales
providing an essential transportation route for local businesses.
The A55 is important locally, nationally and internationally as it forms part of the Euro route
E22 that goes from Holyhead in North Wales to Immingham in North East Lincolnshire.
1.2. Issues Addressed
Road tunnels are inherently dangerous environments and provide the travelling public with
additional risks other than those present on a standard section of carriageway.
There is a requirement under the Road Tunnel Safety Regulations, referring to Annex II of the
EU Directive 2004/54/EC, stating that Automatic Incident Detection (AID) and/or fire detection
must be provided within tunnels that have a greater length than 500m. The project was to
improve the existing incident detection within and around the tunnels and the associated portal
areas of the A55 tunnels in order to improve driver safety and to comply with the Road Tunnel
Safety Regulations.
The original CCTV infrastructure of the North Wales Traffic Management Centre was an
analogue based solution and as such there has been an ongoing general requirement to
upgrade the communications infrastructure to a digital solution; this project has also
inadvertently helped to achieve that requirement.
Incident Detection can be provided using a variety of technologies such as inductive loops and
radar detection; however, both require large amounts of additional infrastructure. The selected
video AID system utilises the existing CCTV infrastructure, thus providing an affordable system
that is compliant with the Road Tunnel Safety Regulations.
The Conwy, Penmaenbach and Pen y Clip tunnels all have a greater length than 500m; the
actual lengths for the three tunnels are as follows:

Conwy tunnel - 1090m;

Penmaenbach tunnel – 660m;

Pen y Clip tunnel – 930m.
The original incident detection system utilised within the A55 tunnels did not provide sufficient
coverage to comply with the EU Directive; this system was only capable of analysing up to six
cameras simultaneously, which is only enough capacity to provide full coverage for one tunnel
bore at any one instance.
Notwithstanding the above, it is also not realistically possible for a small number of control
room operators to monitor CCTV to such a degree that all incidents on the entire network are
immediately detected such that decisive action can be taken prior to the build up of stationary
traffic; this is especially of importance within the tunnel bores. CCTV based AID helps to
eliminate the non-detection of incidents and consequently improves the period of delay in the
EasyWay Evaluation of Video Automatic
Incident Detection
-2-
June 2011
implementation of approved operational and/or emergency procedures currently in place to
protect both the travelling public and the associated tunnel structures alike.
The ability to detect incidents in near real time will assist operators in taking decisive action
such as closing a particular tunnel/bore or appropriate lane(s) accordingly and setting warning
signs and traffic plans as appropriate thereby warning the travelling public of the danger(s)
ahead and/or if required preventing traffic from entering the ‘danger zone’.
An Incident in the tunnels area during busy periods could very easily lead to secondary
incidents and/or fire spread between vehicles. AID assists greatly in the detection period of
such incidents.
EasyWay Evaluation of Video Automatic
Incident Detection
-3-
June 2011
2.
Description of the ITS Project
2.1. Project Domain
The project domains are:

Traveller Information Services;

Traffic Management Services;

ICT Infrastructure.
The table below describes the nature of the A55 tunnel area and the applications that are
affected by this project.
Traveller Information Services
Traffic Management Services
ICT Infrastructure
Variable Message Signs
Automatic Incident Detection
Traffic Monitoring
Driver Behaviour
Use of CCTV
Control Centres
2.2. Objectives
The core objective of the video AID upgrade project was to improve the detection of incidents
within the A55 tunnels, thus reducing the number of secondary incidents and the
corresponding related severity. Secondary objectives of the project were to reduce congestion
caused by such incidents with the subsequent reduction of vehicle emissions during incidents
between junctions 15 and 18. This aligns with the EasyWay objectives i.e. EasyWay is seeking
to achieve an improvement in transport efficiency resulting in a 25% reduction in congestion,
10% reduction in CO2 emissions from transport and to reduce accidents by 25% on the TransEuropean Transport Network (TEN-T) by 2020.
The above stated objectives can be achieved via the rapid identification and notification of
incidents within the A55 tunnels area (J15 to J18) ensuring that appropriate control measures
can be instigated before any relatively minor incident escalates beyond the immediate control
of the tunnel control room operators. The rapid deployment of appropriate control measures
will also aid to ensure that the severity of an incident could be minimised thus ensuring the
number of secondary incidents are also significantly reduced. The minimisation of any
incidents ensures a reduction in congestion with the subsequent reduction of emissions being
achieved.
EasyWay Evaluation of Video Automatic
Incident Detection
-4-
June 2011
2.3. Systems and Technologies Applied
The system configuration for the existing CCTV system for the A55 tunnels can be seen
overleaf in Figure 1. The analogue images from the local CCTV cameras are fed back to
Multiplexers located within the Tunnel Service Buildings (TSBs) and relevant Transmission
Stations (TSs) via a composite video cable. The analogue image is then fed across the fibre
network into De-multiplexers located within the North Wales Traffic Management Centre
(NWTMC). From hereon the image is inputted into the NWTMC CCTV Matrix via composite
video cable, then into a digital encoder. The digital image is then transferred over the IP
network to the visual display units within the Control Room. It should be noted that the PTZ
controls for the cameras are still controlled using the analogue infrastructure through the CCTV
Matrix.
Although multiplexing the CCTV signals into a combined optical signal is a bandwidth efficient
method of supplying wholesale analogue data to a single (remote) location, it is inefficient for
providing images to multiple users and gives rise to resilience issues. Also the equipment
currently utilised to multiplex the analogue CCTV signals is no longer supported by the
supplier. To address the resilience issues with the existing infrastructure and its configuration
the upgraded video AID system was installed to obtain the CCTV images direct from the
cameras as opposed to the previously employed method of receiving them via the NWTMC
CCTV Matrix and associated Multiplexers and De-Multiplexers.
Video AID can operate with both digital and analogue video feeds with equal accuracy.
Although digital signals are post processed analogue signals, when analogue signals are
utilised a video ‘grabber’ is employed to digitize the signal internally. Regardless of the feed,
the actual detection algorithms work using digital image frames (MJPEG). Separate encoders
were installed as part of this project as they provided the following advantages:

Ability to easily connect and control cameras from multiple sites without the need for
additional CCTV Matrices;

Resilience of communications as IP infrastructure is more reliable than other
communication modes;

No bespoke equipment necessary, hardware failures can be rectified using
inexpensive ‘off the shelf’ equipment;

Improved remote monitoring and fault detection.
Additionally, the images were obtained from the analogue CCTV cameras and encoded directly
from the camera to avoid a poor quality image being utilised. There are already a variety of
digital encoders used for the large video screen within the control room and for the web
viewing application. These are housed in the NWTMC which means the image is converted to
digital using an already heavily processed image. Viewing digital feeds in the control room, it is
possible to see that the digital image is affected by this, occasionally causing block pixel errors.
Although not an issue for viewing by sight, this would not be suitable for use with video incident
detection.
EasyWay Evaluation of Video Automatic
Incident Detection
-5-
June 2011
CCTV Camera
(Local Site)
Multiplexer(s)
(TSB/ TS)
De-multiplexer
(NWTMC)
CCTV Matrix
CCTV
Controller
(NWTMC)
(Control Room)
MPEG 4
Encoder for
Video Wall
Control Room
VDUs
MPEG2
Encoder for
AID Server
Web Viewer
AID User
Interface
Figure 1: Existing CCTV and AID Infrastructure
EasyWay Evaluation of Video Automatic
Incident Detection
-6-
June 2011
The upgraded video incident detection system utilises new digital MPEG4/MJPEG encoders
that were installed within the locally situated Tunnel Service Buildings (TSBs) and connected
directly to the camera outputs. The connections are in parallel so that the functionality of the
existing analogue system would not be affected. The new AID analysers were installed within
the TSBs for each of the A55 tunnels for the following reasons:

The analysis software would operate using a cleaner signal with less noise as less
signal processing has occurred at this point;

The video incident detection software must analyse multiple high quality video
channels simultaneously which would consume significant communications bandwidth
if not installed in the locality of the camera;

The video AID system would then be completely independent of the unsupported
communications equipment and the NWTMC CCTV Matrix;

To provide a more resilient system.
A block diagram of how the new AID system was incorporated into the existing CCTV
architecture in order to enable it to operate in parallel can be seen below in Figure 2:
CCTV Camera
(Local Site)
Composite
MPEG4
Video Feed
Encoder (TSB)
Multiplexer
(TSB/ TS)
AID Analyser
Optical Fibre
(TSB)
De-multiplexer
(NWTMC)
Composite
AID Server
Video Feed
(NWTMC)
CCTV Matrix
CCTV
(NWTMC)
Controller
(Control Room)
MPEG 4
MPEG2
Control Room
Encoder for
Encoder for
VDUs
Video Wall
AID
User Interface
Web Viewer
Figure 2: Revised CCTV and AID Infrastructure
EasyWay Evaluation of Video Automatic
Incident Detection
-7-
June 2011
In the event of failure of the video communications equipment or the NWTMC CCTV Matrix,
the digital encoders will provide the operators with the ability to continue to be able to view
CCTV images of the A55 tunnels area in the control room. However, as previously stated, it is
important to note that control of these cameras is not possible via the analogue CCTV user
interfaces using digital communications.
2.4. Status of the Project
Evaluation of this project is ex-post as implementation was completed in July 2010. The
Contract (NEC 3 Option A) was awarded in March 2010 with the works commencing in April
2010.
EasyWay Evaluation of Video Automatic
Incident Detection
-8-
June 2011
3.
Key Evaluation Results
3.1. Impact on Traffic Flow
The subject of this evaluation is the A55 tunnels Video AID Upgrade project. The project
upgraded the incident detection within and around the portals of the immersed dual bore
Conwy tunnel (eastbound and westbound) and the single rock bored Pen y Clip and
Penmaenbach tunnels (both westbound only).
Automatic Incident Detection (AID) can be provided using a variety of technologies of which a
host were considered for this particular site. After much deliberation and review of the available
technologies it was decided that a Video AID system based on the utilisation of existing Pan,
Tilt and Zoom (PTZ) type CCTV cameras and the associated infrastructure be deployed for
AID for the three tunnels on the A55 in North Wales. A complete renewal of the existing
infrastructure was dismissed as it was deemed too expensive.
Video based AID provides traffic operators with a system that can detect stopped vehicles
within a carriageway by utilising video signals from any CCTV camera that is connected to the
network. The type of camera is irrelevant as they could be of the fixed type, PTZ, analogue,
Internet Protocol (IP) or any other type thus requiring no additional hardware or cabling to be
installed or integrated into the existing infrastructure within the tunnel bores.
Utilising CCTV cameras as opposed to other technologies (e.g. induction loops) provides a
vast monitoring area for all directions and traffic lanes that are within camera view. The system
provides traffic operators with both an audible and visual alarm within a few seconds after an
incident (e.g. stopped vehicle) occurs, even before the consequences of an incident can be
noticed by more traditional monitoring techniques.
The chosen supplier for the video AID upgrade system was Citilog, a French company. The
software application is known as VisioPaD. VisioPaD enables the operating authority to change
an existing video surveillance system into an advanced video surveillance system. The system
works in the ‘background’ on all integrated video images thereby allowing the control room
operators to focus on other tasks. VisioPaD is part of the Media range of software products
which includes MediaTunnel, MediaServer and MediaRoad. VisioPaD has a reduced set of
functions in comparison to the above mentioned products but is unique in that it is a self
learning system designed to work specifically with PTZ cameras.
As stated above, the chosen AID software application is specifically designed to operate on
PTZ cameras; it automatically and dynamically self-adapts to camera movements and it does
not require any setup or calibration. It covers all lanes of traffic within its view and certain
detection and data collection settings are fully adjustable. The system runs on standard
hardware and standard IP communication protocols and is based on a client-server
architecture with a user friendly interface. A database facility for the storage of video clips for
post incident analysis is also provided.
EasyWay Evaluation of Video Automatic
Incident Detection
-9-
June 2011
The VisioPaD algorithm extracts information on objects inside the monitored area on a
succession of pictures from CCTV cameras. At application start-up, the system creates a
background image that is permanently updated. The presence of vehicles is detected in each
image by comparison between the actual image and the ‘background’ image. The algorithm
tracks the vehicle through the sequence of images and analyses its movement to build up the
time-space trajectory. Any movement of the camera (Pan, Tilt or Zoom) is immediately
followed by a self learning process performed on the ‘new’ field of view. It takes approximately
one minute to re-evaluate the ‘background’ at which point it again becomes operational for the
detection of stopped vehicles.
3.2. Impact on Safety
The video AID system improves driver safety by providing real time automated incident
detection enabling the possibility of deploying corrective mitigation measures before losses
escalate or traffic congestion builds up to unacceptable proportions. An additional benefit of the
system is that it provides network managers with a facility that will also identify any
unauthorised access to sensitive structures that are on the network e.g. tunnels.
The employment of the new video AID system and the associated additional hardware that was
subsequently required to provide IP compatible digital images to the new AID analysers has
had an additional benefit of providing resilience to the existing CCTV infrastructure such that
there is now built in contingency for the viewing of the CCTV images in the NWTMC in the
event of existing analogue infrastructure faults and/or failures.
3.3. Impact on Environment
The video AID system provides additional benefits of helping with the reduction of vehicular
emissions during an incident. The larger the size of an incident or the longer the incident lasts
the more harmful pollutants that are emitted into the atmosphere. The video AID system
indirectly supports governments with their objectives of reducing the overall pollutant
emissions, by ensuring incidents are noticed or identified immediately thereby minimising the
size and the length of any incident.
Due to the utilisation of existing cameras and the associated existing infrastructure there have
only been minimal additional power requirements for the additional hardware (video analysers).
The new hardware has been installed in existing buildings associated with the network
ensuring the system has a low carbon footprint.
3.4. Other Key Results
Other indirect benefits of the video AID system are reduced driver stress and improved driver
comfort. The driving experience is expected to be enhanced due to the removal of long delays
caused by tunnel incidents not being immediately detected and subsequently dealt with; this
will ensure that appropriate mitigation measures are immediately employed.
The improved detection of incidents within the A55 tunnels will also reduce the number of
secondary incidents (e.g. rear end shunts) caused by queuing traffic; this is achieved by
displaying suitable warning messages on the electronic signs located on the tunnel
approaches, thus improving safety to the travelling public. The reduction in secondary incidents
will reduce the overall cost of an incident and reduce the demand on the emergency services.
EasyWay Evaluation of Video Automatic
Incident Detection
- 10 -
June 2011
4.
Evaluation Planned
4.1. Timing and Type of Evaluation
The upgraded system was installed in July 2010. An ex-post evaluation has therefore been
completed reviewing 10 months of data that has been obtained from the video AID system
database that has stored video captions. These video captions are all those ‘incidents’ (both
genuine and false alarms) that the AID server has determined to be classed as such and
therefore captured and stored on a SQL database on the server.
4.2. Objectives for the Evaluation
The evaluation objectives are to identify the degree to which the video AID system contributes
in the achievement set by EasyWay. This can be achieved by completing the vendor’s
Performance Computation Methodology to obtain an accurate understanding on how the
system is operating.
4.3. Research Questions
The following points were the main topic of this review of the video AID system for the A55
North Wales tunnels:

Efficiency;

Effectiveness;

Operator use-ability, procedures and perception;

Flexibility;

Value for money;

Possible Improvements.
4.4. Study Area for the Evaluation
The study areas used for the ex-post evaluation of the new video AID system are the thirty
CCTV cameras that the video AID analysers have been connected to. The cameras located
within the Conwy, Penmaenbach and Pen y Clip tunnels are located on the A55 between
junctions 15 and 18. The area in concern is a stretch of the A55 that is approximately 11
kilometres in length. It must be noted, however, that full coverage is not available of the
complete stretch of road.
EasyWay Evaluation of Video Automatic
Incident Detection
- 11 -
June 2011
4.5. Expected Impacts
The expected impact of the video AID system is to reduce congestion caused by incidents
through improved reaction times by the tunnel system operators, to improve safety of the
tunnel structures, reduce vehicle emissions during such incidents and to improve the motorists’
general welfare and safety on the associated section of the network under continual AID
monitoring.
The new video AID system provides the above benefits by providing real time automated
incident detection that enables corrective mitigation measures to be deployed prior to any
incident escalating beyond intolerable levels. Once an incident has been detected by the
system, the control room operators are alerted via an audible and visible alarm. The operator
is required to acknowledge the incident and confirm whether it is a false alarm or an actual
incident; there is also a facility to add a short note and description for future scrutiny. A copy of
the video/incident is then kept on the video AID server SQL database for future use.
Once an actual incident has been acknowledged and subsequently logged, the operator,
following approved operational procedures deploys the correct mitigation procedure
appropriate to the incident. Mitigation measures can be the setting of appropriate warning
messages on the upstream signs available within the area to ensure that the oncoming
motorists are forewarned of the incident ahead and/or measures could include the deployment
of Traffic Officers to the incident. Dependant on the severity of the incident the operator can
also instigate emergency lane closures or a full tunnel closure (under instruction from the
relevant authority) via a separate Variable Message Signing System (VMSS) thus ensuring that
the incident does not escalate beyond control. The operator also has the use of an
Environmental Control System (ECS) within the tunnel providing ventilation and/or lighting
controls that may assist with the welfare of the travelling public and/or the emergency Services
within the tunnel bore.
By taking these appropriate mitigation measures the impact and severity of any incident can be
significantly reduced therefore ensuring that any knock on affects such as secondary collisions
and congestion with the associated increased pollution are reduced. The reduction in
congestion will ensure that the welfare of the motorist is improved by reducing delays caused
by incidents whilst reducing the number of secondary incidents will reduce the overall size of
the incident. Vehicle emissions will also be reduced due to the reduction in congestion and
impact of the incident.
The video AID system also improves the safety of the tunnel structures by detecting if any
unauthorised vehicles/persons have stopped, or are located within their vicinity. The system
will again alert the operators of any vehicles that have been detected, allowing them to take the
appropriate action to ensure the structure remains safe.
Most importantly is the fact that the video AID system improves the safety of the tunnel
structure and the users (travelling public) by providing the network control room operators with
a system that alerts the operator to an incident in real time thus enabling suitable mitigation
measure to be rapidly deployed as it is extremely important to ensure that the number of
vehicles involved in an incident are kept to an absolute minimum during major incidents.
EasyWay Evaluation of Video Automatic
Incident Detection
- 12 -
June 2011
5.
The Impact of the Project - Results
5.1. Technical Performance
This section describes the technical performance of the system, and provides conclusions
about the influence of technical performance on the impacts identified.
5.2. Results
An evaluation of the video AID system has been carried out using data that has been collected
and stored on the AID SQL database server, 10 months of stored data has been reviewed as
part of the evaluation. Details of the results are contained below:
Conwy Tunnel Westbound – Video AID
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
28
13
65%
21
39
35%
24
10
68%
36
18
65%
27
9
71%
21
(Conwy W-1)
22
(Conwy W-2)
23
(Conwy W-3)
24
(Conwy W-4)
25
(Conwy W-5)
Conwy Tunnel Eastbound – Video AID
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
72
13
84%
70
8
88%
49
18
68%
71
(Conwy E-1)
72
(Conwy E-2)
73
(Conwy E-3)
EasyWay Evaluation of Video Automatic
Incident Detection
- 13 -
June 2011
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
29
14
67%
48
6
87%
74
(Conwy E-4)
75
(Conwy E-5)
Pen y Clip Tunnel – Video AID
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
25
48
34%
25
16
60%
46
37
55%
24
23
51%
17
21
45%
23
10
70%
41
(Pen y Clip 1)
42
(Pen y Clip 2)
43
(Pen y Clip 3)
44
(Pen y Clip 4)
45
(Pen y Clip 5)
46
(Pen y Clip 6)
Penmaenbach Tunnel – Video AID
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
30
27
53%
48
100
32%
26
10
72%
40
42
49%
31
(Pbach -1)
32
(Pbach – 2)
33
(Pbach -3)
34
(Pbach -4)
EasyWay Evaluation of Video Automatic
Incident Detection
- 14 -
June 2011
Outside Cameras – Video AID
Camera No.
No. of
Detections
No. of False
alarms
Detection Rate (%)
635
108
85%
108
123
47%
280
27
91%
76
(Conwy E-Portal)
77
(Conwy E-Portal)
47
(Pen y Clip WPortal)
During the evaluation period a number of non-detections were logged by the system
administrator, details of which can be seen below:
Date of Incident
Time of Incident
Description
Camera Number
24/05/2011
02:22
Conwy E2
72
15/05/2011
17:41
Conwy E3
73
05/05/2011
12:41
Conwy E3
73
11/03/2011
18:11
Conwy W Portal
70
04/02/2011
14:48
Conwy E3
73
13/01/2011
19:07
Conwy W5
25
12/01/2011
20:40
Conwy E3
73
11/01/2011
21:54
Conwy E3
73
09/11/2010
18:05
Conwy E5
75
19/10/2010
18:01
Conwy W Portal
70
EasyWay Evaluation of Video Automatic
Incident Detection
- 15 -
June 2011
As can be seen from the tables above there are a certain number of specific cameras that
experience relatively low detection rates in comparison to other camera locations i.e. the rate
of false alarms compared to actual stopped vehicles is considerably higher. This is due to
various reasons such as the actual image quality being poor, the camera set-up and
environmental impacts such as light refraction/reflection, wind and rain.
5.3. Reliability of Results
The results depicted in the table in Section 5.2 above are those instances of activation of the
video AID system as logged and stored on the AID database server. The reliability of the
results assessed as part of this review can not really be judged or calculated as those events
that have been logged on the server are a complete record of all activations of the video AID
system.
The last section of the table is those instances of incidents (stopped vehicles) that the video
AID did not detect. It is the opinion of the author that these ‘non-detections’ are far more
important than any issues associated with false alarms. However, saying that false alarms are
in themselves detriment to the effectiveness of a system as a whole because the system
operators in general have a tendency to ignore alarms if the said system is constantly
generating spurious alarms.
5.4. Research Questions Answered
With reference to the research questions set out before the evaluation started (summarised in
Section 4.3), the following section describes how the evaluation provided answers to the
questions posed.
5.4.1. EFFICIENCY
The video AID system is a very useful tool in the aid of running a control room with limited staff
that also have the responsibility for the monitoring and control of multiple tunnel and road
based control systems. The video AID system allows the control room operators to concentrate
on the various tasks without the constant requirement for the stringent monitoring of the CCTV
cameras in and around the A55 tunnels area.
5.4.2. EFFECTIVENESS
As can be seen by the results table in Section 5.2 the video AID system is relatively good at
detecting stopped vehicles as long as those images available to it are of good quality with
limited interference of the image quality such as environmental influences (wind/vibration), rain,
lighting changes and object masking by other passing vehicles.
The video AID system, due to its nature of automatically detecting stopped vehicles can lead to
operator complacency where the operators tend to not pay as much attention to the real time
CCTV images as they tend to become reliant on the AID to detect incidents on their behalf and
alert them accordingly as and when required. This is however, not quite the case in this
instance as the video AID software application (VisioPaD) is not as effective as it possibly
could be at detecting stopped vehicles. This is not necessarily a criticism of the video AID
application but is more an acknowledgement of the fact that there are limitations as to how
effective a system like this can be on PTZ cameras and the existing infrastructure and quality
of existing CCTV cameras. Inevitably it must be accepted that video detection is not an exact
EasyWay Evaluation of Video Automatic
Incident Detection
- 16 -
June 2011
science and this technology, especially for PTZ cameras, is in its infancy and is continually
being developed and improved.
Another rather pertinent issue with the system is its inability to deal with rapid environmental
changes and inclement weather. This is particularly of importance to those cameras that are
located in areas subject to the elements i.e. camera locations at the perimeters of the tunnel
structure get affected by ‘sudden’ light changes and cameras physically outside of the tunnel
bores are constantly affected by rain and wind.
5.4.3. OPERATOR USABILITY AND PERCEPTION
The video AID system client software user interface is easy to use and is simple to navigate
around the various facilities provided for general use. As stated above though the video AID
system can not be 100% guaranteed to detect all instances of stopped vehicles due to various
reasons. This therefore leads to the false assumption that the software is not functioning
properly and criticism is then directed at the system.
This is not really the case either and the operator and/or operating authority needs to accept
his/their part in the overall effective operation of the video AID system as a whole. It must be
appreciated that the system can only process images that it is presented with. Therefore, the
operators need to continually assess whether the image presented for manipulation is the best
quality and is fit for accurate and effective digital signal processing purposes e.g. if the camera
is zoomed out too far, it is quite unfair to expect the system to pick up what would be a
relatively small ‘blob’ stopping in the background. Similarly, and especially towards the tunnel
extremities there is a lot of light glare and lighting reflections which also cause the system to
miscalculate which can lead to non-detections of stopped vehicles and increased false alarms.
It is therefore of the highest importance that the system operators are made aware of the
limitations of the system and for their (the operators) requirement to be pro-active and
dynamically alter the view of the cameras in accordance with the actual surroundings and the
environmental conditions at all times in order to achieve the optimum results from a PTZ based
video AID system.
Notwithstanding the above, a possible solution for improvement to the existing system is
presented in Section 5.4.6
5.4.4. FLEXIBILITY
The AID system is fairly flexible in that the sensitivity settings can be configured and ‘tweaked’
for optimum performance based on the existing infrastructure. Additional cameras can also be
added to the system without the requirement for additional hardware but at an added cost for
licenses i.e. each camera input requires a separate software license.
EasyWay Evaluation of Video Automatic
Incident Detection
- 17 -
June 2011
5.4.5. VALUE FOR MONEY
The new NWTMC video AID system that was employed for the A55 tunnels is considered as
being very good value for money as it could be employed on the existing CCTV infrastructure
with minimal cost for supplementary hardware. The added cost of replacing the PTZ cameras
with fixed cameras would have been significant; in addition the total camera count would have
to be increased to give the same coverage as can be obtained with camera zoom (PTZ). This
would have therefore, made it uneconomical to provide a video AID for the A55 tunnels which
would have led to increased risks for the operating authority, and to the road users, as well as
not complying with the Road Tunnel Safety Regulations, Annex II of the EU Directive
2004/54/EC, stating that Automatic Incident Detection (AID) must be provided within tunnels
that have a greater length than 500m.
5.4.6. POSSIBLE IMPROVEMENTS
The video AID vendors are currently in the process of developing system software that can
implement their normal AID application designed for fixed cameras on PTZ cameras when the
PTZ type cameras are in their ‘home’ or ‘preset’ position. This software allows the user to draw
‘masks’ on the image of the road infrastructure thereby eliminating all other outside influences
from any data manipulation; this would obviously be a far more effective method of providing
video AID as this method of providing video AID is relatively advanced and the algorithms
associated with this method of detection are tried and tested.
As part of the vendors’
maintenance contract options, a minimum of one software upgrade per annum is provided and
it is proposed to recommend this option to the client as the new software mentioned above will
be provided under this maintenance contract. This will ensure that the NWTMC’s video AID
system remains ‘state of the art’ while the ongoing cost to the operating authority is minimised.
Further improvements available to the overall CCTV system as a whole would be the upgrading
of the existing analogue cameras and associated infrastructure to a full digital solution, this
would certainly improve the quality of the images presented to the video AID system for
manipulation.
5.5. Overall Assessment
This section summarises the main findings in terms of their impact on European objectives
under the headings below.
5.5.1. SAFETY
The safety of the travelling public has been increased due to the incorporation of the video AID
system.
5.5.2. EFFICIENCY
As stated before the addition of the new video AID has proven to be a great asset to the overall
efficient running of the North Wales Traffic Management Centre control room. The detection
rate of stopped vehicles and/or incidents within the tunnels area on the A55 has also improved
since the introduction of the system.
EasyWay Evaluation of Video Automatic
Incident Detection
- 18 -
June 2011
The upgraded system is still in its first year of employment and it does require some subtle
configuration changes (user configurable and camera) in order to obtain optimum performance
levels. This may cause an increase in the rate of false alarms on the affected cameras but
should be attempted and suitably investigated/assessed. However, as previously stated too
many false alarms can be detrimental to the effective use of the system due to system
operators becoming complacent and subsequently ignoring alarms.
The vendors’ upgrade options (via their maintenance contract(s)) should also be carefully
considered as it is perceived that this will improve the detection rates significantly and at the
same time it should also reduce the amount of false alarms. Regarding the current systems
efficiency when utilised under inclement weather conditions it is considered that those cameras
that are severely affected under these conditions i.e. those that are outside of the tunnel bores
and not protected from rain and wind vibration etc. should be disabled from the system until the
environmental conditions improve as they will constantly provide false alarms under these
conditions.
5.5.3. ENVIRONMENT
From an environmental aspect the project was a complete success as the existing CCTV
cameras and associated infrastructure was employed. Due to the fact that a very limited
amount of additional hardware was required to successfully implement this project, the
additional power requirements and maintenance thereof is minimal. The new video AID system
therefore has a very small carbon footprint.
5.6. Additional Aspects
The upgraded software, available via the above mentioned maintenance contract will also allow
smoke detection as well as provide data analysis facilities on traffic flow. The detection of
smoke is vital for the safe operation of road tunnels and the addition of this capability is
paramount to the future running and compatibility to any existing and future standards
applicable to the three road tunnels on the A55.
The traffic monitoring features of the upgraded software should also prove to be very useful for
network stakeholders and can be manipulated for the general day to day management and any
future planned traffic modelling of the A55 and the associated trunk roads network.
EasyWay Evaluation of Video Automatic
Incident Detection
- 19 -
June 2011
6.
European Dimension: Transferability of the Results
Video AID can be used on any CCTV installation; the results of this report are therefore
transferrable to any member state or organisation within the EU.
This section provides a breakdown of the costs and the methods employed for the evaluation
of the video AID product.
Annex 1: Technical Annex
6.1. Breakdown of Project Costs
The capital costs for the procurement of both the software and hardware and the installation
costs for the video AID system are detailed below;
ITS Technologies
Licences, software, etc
€124,425
£112,500
€45,346
£41,000
€56,959
£51,500
€226,730
£205,000
ITS Infrastructure
Hardware and other ancillary items
Installation Cost – Technologies & Infrastructure
Configuration, training, support and warranty
Total
As discussed previously, the video AID system vendors offer yearly maintenance packages
that include software upgrades in addition to continual fine tuning of the system. Support cost
for the first 12 months after installation was included within the above costs. Continual support
is recommended to ensure that the system is operating within its desired parameters and to
minimise the amount of false alarms and non-detections.
The vendors offer four maintenance packages with varying level of support and services
ranging from remote diagnostic checks to annual software upgrades and site visits.
EasyWay Evaluation of Video Automatic
Incident Detection
- 20 -
June 2011
6.2. Data Collection Methods
PERFORMANCE COMPUTATION METHODOLOGY
1.
INTRODUCTION
1.1
Purpose of Document
The purpose of the document is to present all performance evaluation criteria for an AID
system, as well as methodology elements regarding this evaluation.
2.
DEFINITION
The performance level of AID systems is measured by three indicators that are not
independent:
2.1
Averaged Detection Time
The detection time measures the reactivity of the system. For stop detection, it is an adjustable
parameter for the AID Administrator.
Detection time is defined as the time interval between an event and its appearance on the
server.
2.2
Detection Rate
The detection rate measures the capability of the system to properly detect specified incidents,
as described along this document.
DETECTION RATE 
total number of det ected event
total number of specified event
The detection of an incident does not take the following information into account:

The lane of detection.
o

If the detection is made on an adjacent lane, the detection is correct.
The exact position of the marker on the image.
o
The position of the red arrow may not be exactly on the object to be detected
for various geometrical reasons.
In a later stage, it is possible to monitor the type of the alarms. Please note that the type of the
alarms corresponds to an estimation of the algorithm. For instance:

A STOPPED VEHICLE can be seen as DEBRIS (and vice versa) if the size estimation
is wrong or difficult to perform (far in the field of view).

More rarely, a STOPPED truck can be seen as LOSS OF VISIBILITY (and vice versa)
because a truck may obstruct a large portion of the image.
Upon project realization, the true detection of event (which corresponds to an alert to the
operator) is more important than the type (additional information to the operator that may be
verified visually). For example a detection of SLOW VEHICLE seen as a PEDESTRIAN will be
counted as a good detection.
2.3
False Alarm Frequency
The false alarm rate measures the system reliability. A false alarm occurs when the system
sends an alarm whereas no event that should be detected is occurring in the image. It is
EasyWay Evaluation of Video Automatic
Incident Detection
- 21 -
June 2011
expressed in number of false alarms per camera and per day. This value measures the
annoyance induced to the operator by the system.
The false alarms can be classified in two categories:
External false alarms can be solved by improving the video chain without modifying the AID
system. For example:

False alarms due to bad image quality;

False alarms due to rain on the camera window;

False alarms due to vibration whose amplitude is out of specifications;

False alarms due to lighting whose changes are out of specifications.
Internal false alarms correspond to the system limits. The vendor developed methodologies to
limit those false alarms, knowing that a zero false alarm system does not exist.
Possible causes of false alarms are the same oftentimes, and correspond to situation difficult
to analyze for a video-based AID system:

Structured shadows in outdoor environment (cables in the case of a bridge);

Non-recurring lighting changes;

Wheel marks on wet road;

Presence of objects on the roadway (paper, puddle).
Experience showed that less than 20% of total internal false alarms do not belong in one of the
stated categories. This means that most of the false alarms occur for one or more of the stated
reasons, and thus on specific cameras, typically cameras close to tunnel extremities.
As a consequence, the false alarm rate must be given for a full set of cameras, including
cameras with very good performances and harder cameras.
EasyWay Evaluation of Video Automatic
Incident Detection
- 22 -
June 2011
Download