PPR434 Test of TrafiCam vehicle detector in Winchester K Wood
Transport Research Laboratory PUBLISHED PROJECT REPORT PPR434 Test of Traficam vehicle detector in Winchester by K Wood (TRL) Prepared for: Project Record: PPRO 04/16/17 Client: Emerging detector technologies Department for Transport, Traffic Management Division (Suku Phull) Copyright Transport Research Laboratory August 2009 This Published Report has been prepared for the Department for Transport. Published Project Reports are written primarily for the Client rather than for a general audience and are published with the Client s approval. The views expressed are those of the author and not necessarily those of the Department for Transport. Name Date Approved Project Manager A Kirkham 19/08/2009 Technical Referee I J Burrow 19/08/2009
When purchased in hard copy, this publication is printed on paper that is FSC (Forestry Stewardship Council) and TCF (Totally Chlorine Free) registered. TRL PPR434
Contents Executive summary i Abstract 1 1 Introduction 1 2 Test site and installation 3 3 Data collection and analysis 6 3.1 SCOOT data collection 6 3.2 Data analysis 7 4 Results 9 5 Conclusions 13 Acknowledgements 13 TRL PPR434
TRL PPR434
Executive summary The Department for Transport commissioned TRL to assist in the development of vehicle and pedestrian systems for urban traffic management. One of the strands of the work is to assess the performance of new detection systems. Hampshire County Council kindly provided a test site in Winchester where trial detectors could be mounted on a lamp column adjacent to a SCOOT loop and the output from the trial detector returned to the SCOOT computer for off-line comparison with the output of the inductive loop detector. A TrafiCam video detector was installed on the lamp column and connected to the SCOOT computer via the Out-station Transmission Unit in the signal controller. The output from the TrafiCam detector was configured in SCOOT to be associated with a dummy link that had no effect on the optimisation of the signal timings. Associating the detector with a SCOOT link allowed the use of SCOOT messages to monitor the output of the detector. Standard SCOOT messages giving the status (off or on) of the TrafiCam and the corresponding SCOOT loop were recorded and analysed. The conclusions from the analysis were: The overall agreement between the TrafiCam and the inductive loop was good. The TrafiCam detection system tested for SCOOT seemed to work well in comparison with a loop system. Because of the need for the TrafiCam to see the rear lights of vehicles, its detection zone was not coincident with the inductive loop. Therefore, comparison of individual vehicle counts in slow moving traffic was more difficult than it would have been with coincident detection zones. Nevertheless, 99% of individual vehicle detections were matched. Total occupancy recorded by the TrafiCam was about 10% greater than the loop. Some of this difference could have been due to a slightly longer effective detection zone. In addition, the TrafiCam detection zone was downstream of the loop and, therefore, more likely to be queued in, resulting in more vehicles being delayed in its detection zone, hence greater occupancy than over the loop. The SCOOT validation process would allow for a small difference in the length of the detection zone, which would then have no practical effect on the operation of SCOOT. There were some differences in the outputs of the two detectors when traffic was queuing back to the detector zones. Some of the differences could be explained in terms of the difference in detection zones, but in a very few cases it appeared that the TrafiCam might have stuck on for some time (up to half a minute) during which the loop detected several vehicles. TRL i PPR434
TRL ii PPR434
Abstract A TrafiCam video vehicle detector was installed adjacent to a standard SCOOT inductive loop detector in Winchester. Outputs from the two detectors were returned to SCOOT and recorded for analysis. The inductive loop output was used in the SCOOT control algorithms as normal, but the trial TrafiCam detector was ignored by them. The status of both detectors, off or on, each quarter second was output in standard SCOOT messages and compared. The comparison was complicated as it was necessary to define the TrafiCam detection zone slightly downstream of that of the loop, but it was possible to make valid comparisons. 1 Introduction The Department for Transport commissioned TRL to assist in the development of vehicle and pedestrian systems for urban traffic management. One of the strands of the work is to assess the performance of new detection systems. Hampshire County Council kindly provided a test site in Winchester where trial detectors could be mounted on a lamp column adjacent to a SCOOT loop and the output from the trial detector returned to the SCOOT computer for off-line comparison with the output of the inductive loop detector. TRL 1 PPR434
TRL 2 PPR434
2 Test site and installation The test site is on the Romsey Road in Winchester at its junctions with Battery Hill and Kings School, see Figure 2-1. Figure 2-1: Winchester test site The lamp column is conveniently situated opposite the loop for traffic travelling south away from Winchester, ideal for testing side-fire detectors, and the signal controller is behind the bushes. However, the TrafiCam does not operate in true side-fire as the best performance for night-time operation is to view the rear lights of vehicles. Therefore, the detector was mounted on the lamp column looking slightly downstream of the loop. The resulting difference in detection times, TrafiCam activated slightly after the loop, was allowed for in the analysis of the results. The mounting bracket was attached to the lamp column by metal bands and the camera detector attached to the bracket using the standard short metal rod, see Figure 2-2. The camera lens is difficult to discern in the photo, but is a small circle about 2 cm diameter and the camera enclosure is 10 cm in diameter. Details of the unit are available on the web site http://www.traficam.com/home.jsp At the test site, the unit was installed at between 4 and 5 metres high, giving an adequate view for one lane of traffic. A junction box connected by armoured cable to the controller had been installed at the site and the TrafiCam was wired into this box. TRL 3 PPR434
Figure 2-2: TrafiCam The interface card was installed in the controller and connected to the feeder cable. The camera communicates to the interface card using RS485. A laptop PC running the custom set up software was connected to the interface card and a camera image downloaded to it. The required detection zone was drawn on the image. After the detector and interface had been fully configured the detector was initiated. Operation of the unit was checked both through the interface and by comparing the operation of the LED on the camera housing with vehicles passing through the zone. The final verification of the performance was performed by recording the operation of the detector and a video image of the detection zone for several hours on the PC for later review. The detector had been set up in daylight, but the Traficon engineer was keen to check that the parameters for night-time detection were set correctly to operate on tail lights with the site ambient lighting. Figure 2-3 and Figure 2-4 show two screen shots from the PC TRL 4 PPR434
with the detection zone shown in green. The arrow shape at the top of the zone is an indicator that the detection zone is direction sensitive. Only vehicles travelling in that direction, or stopped vehicles that were travelling in that direction, are detected. Figure 2-3: Daylight screen shot TRL 5 PPR434
Figure 2-4: Night time screen shot 3 Data collection and analysis 3.1 SCOOT data collection The TrafiCam interface card in the controller was connected to the Outstation Transmission Unit (OTU) and the detector output returned to the central SCOOT controller together with those of all the other detectors connected to that controller. The SCOOT data was configured so that the TrafiCam output was associated with a dummy link at the same node as the nearby SCOOT loop. That link was monitored by SCOOT, but not used by any of the optimisers and so it had no influence on the signal operation. However, as a fully configured SCOOT link all the SCOOT link and detector based messages were available. In particular the M19 message that gives details of the detector state could be recorded for analysis. In a SCOOT system the OTU samples each detector every quarter of a second to take a snapshot of its status, on in which case a 1 is sent to SCOOT, or off when a 0 is sent. For each detector, four bits indicating the detector status for the last second are returned to SCOOT every second. These four bits can be output in the M19 message. An example is shown below for data recorded on the 18 th February: ****** Start of SCOOT log for 18-FEB-09 ****** We 12:00:00 M19 N42121B1 DETECTOR STATE 0000 We 12:00:01 M19 N42121B1 DETECTOR STATE 0000 We 12:00:02 M19 N42121B1 DETECTOR STATE 0000 We 12:00:03 M19 N42121B1 DETECTOR STATE 0000 TRL 6 PPR434
We 12:00:04 M19 N42121B1 DETECTOR STATE 0000 We 12:00:05 M19 N42121B1 DETECTOR STATE 0000 We 12:00:06 M19 N42121B1 DETECTOR STATE 0000 We 12:00:07 M19 N42121B1 DETECTOR STATE 0000 We 12:00:08 M19 N42121B1 DETECTOR STATE 0000 We 12:00:09 M19 N42121B1 DETECTOR STATE 0000 We 12:00:10 M19 N42121B1 DETECTOR STATE 0000 We 12:00:11 M19 N42121B1 DETECTOR STATE 0000 We 12:00:12 M19 N42121B1 DETECTOR STATE 0000 We 12:00:13 M19 N42121B1 DETECTOR STATE 0111 We 12:00:14 M19 N42121B1 DETECTOR STATE 0000 We 12:00:15 M19 N42121B1 DETECTOR STATE 1110 We 12:00:16 M19 N42121B1 DETECTOR STATE 0001 We 12:00:17 M19 N42121B1 DETECTOR STATE 1000 We 12:00:18 M19 N42121B1 DETECTOR STATE 0000 We 12:00:19 M19 N42121B1 DETECTOR STATE 0000 We 12:00:20 M19 N42121B1 DETECTOR STATE 0000 A message is output each second for each detector for which the message is turned on. The format of the message is: Two letters to denote the day of the week, in this example, Wednesday The time, hh:mm:ss The message identifier (M19) Detector reference number (node N42121, link B, first detector on the link) Detector state in the previous second starting with the first quarter second, 0=off, 1=on In the 21 seconds shown above three vehicles were detected: Vehicle 1 detected in the last 3 quarter seconds of the 12:00:13 message Vehicle 2 detected in the first 3 quarter seconds of the 12:00:15 message Vehicle 1 detected in the last quarter second of the 12:00:16 message and the first quarter second of the 12:00:17 message The M19 message was logged for the TrafiCam detector and the neighbouring inductive loop. 3.2 Data analysis SCOOT uses both the flow of vehicles (a new vehicle is indicated by an off to on transition in the detector status) and the length of time each vehicle activates the detector (occupancy for each vehicle). Therefore, the data analysis was designed to analyse the data in detail, to count individual discrepancies between detection of a vehicle by each detector rather than total flow in an hour, for example. As noted in the description of the site, the detection zones for the loop and TrafiCam were not coincident due to the need of the TrafiCam to view the rear of vehicles for night-time detection. The distance between the detection zones was constant, but the time between detection in the two zones was not. At busy times, queues build back towards the junction from the next junction downstream, slowing vehicles in the vicinity of the detectors. Therefore, at some times the start of detection of a vehicle at the inductive loop will be well in advance of detection by the TrafiCam. At other times the difference in initial detection will be much shorter. TRL 7 PPR434
Under free-flow conditions, the difference in detection times was less than a second. Therefore, a vehicle detection by the TrafiCam within one second of a detection by the loop was classified as a successful detection by both detectors. However, a second criterion was needed for congested conditions. A vehicle could be delayed by the queue and remain stationary, detected by the loop for some time before moving into the detection zone of the TrafiCam. However, when it started to move, it would be expected to enter the TrafiCam s detection zone soon after leaving that of the inductive loop. Therefore, the second criterion for successful detection was defined to be detection by the TrafiCam within half a second of ceasing to be detected by the loop. Three categories of TrafiCam detection were defined: 1. Successful detection: Start of detection by the TrafiCam within 1 second of the start of detection by the inductive loop Or start of detection by the TrafiCam within ½ second of the end of detection by the inductive loop 2. False positive: Detection by the TrafiCam without a detection by the inductive loop meeting either of the above criteria 3. False negative: Detection by the inductive loop without a detection by the TrafiCam meeting either of the above criteria In addition to counting successful detections, false positives and false negatives, the total occupancy and occupancy per vehicle were calculated. TRL 8 PPR434
4 Results The overall results are summarised in Table 4-1. The matches, false positives and false negatives are as defined above. In congested conditions it is possible that the time lag between a small vehicle leaving the inductive loop and entering the TrafiCam detection zone could have been greater than 0.5 seconds, resulting in both a false negative and a false positive being recorded. Therefore, the match rate may be slightly higher than the 99.0% quoted in the table. The column headed ratio is the ratio of the TrafiCam result to the relevant inductive loop data, e.g. the false positives amount to 0.9% of the vehicles counted by the inductive loop. Table 4-1: Summary results Inductive loop TrafiCam Ratio Vehicle count (vehicles) 23333 23323 99.96% Occupancy (s) 14911 16331 110% Matches 23102 99.0% False positives 221 0.9% False negatives 231 1.0% These results are generally good and give confidence that the TrafiCam would normally provide good information to SCOOT. The detector occupancy is noticeably higher for the TrafiCam than the loop, which is not intrinsically a problem as some latitude is allowed in the length of the detection zone and catered for in the SCOOT validation process. However, examining the data in detail showed that there were occasions when the TrafiCam appeared to stick in the on state compared with the loop. There was a suspicion that on some occasions the appearance of a new vehicle, according to the loop, caused the TrafiCam to resume normal operation. It was noted that the inductive loop appeared to have an upper limit on its occupancy of 3.5 seconds in most of the data collected. The presence time of the detector should be much longer than this to register congestion and allow SCOOT to respond. It was not possible from the data to be sure that there was not a problem with the loop causing problems with comparison of the detectors in congested conditions. However, in some cases where the TrafiCam was on for considerably longer than the loop it was possible to rationalise the discrepancy in terms of the difference in detector zone along the road. The TrafiCam detection zone was further from the upstream junction, and therefore closer to the downstream junction, than the loop. Therefore, queues from the downstream junction were more likely to affect the TrafiCam detection zone than the loop. An example below shows vehicles highlighted on the two detectors. It appears that the queue just reaches to the loop, but it delays some vehicles more slightly downstream in the TrafiCam detection zone. The third vehicle is detected for 6 seconds by the TrafiCam, when that one clears the detection zone, the fourth vehicle leaves the loop, coincidentally after 3.5 seconds of occupancy. TRL 9 PPR434
Loop Tu 11:55:23 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:24 M19 N42121B1 DETECTOR STATE 1110 Vehicle 1 Tu 11:55:25 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:26 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:27 M19 N42121B1 DETECTOR STATE 0111 Vehicle 2 Tu 11:55:28 M19 N42121B1 DETECTOR STATE 1110 Tu 11:55:29 M19 N42121B1 DETECTOR STATE 0111 Vehicle 3 Tu 11:55:30 M19 N42121B1 DETECTOR STATE 1111 Tu 11:55:31 M19 N42121B1 DETECTOR STATE 1100 Tu 11:55:32 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:33 M19 N42121B1 DETECTOR STATE 1111 Vehicle 4 Tu 11:55:34 M19 N42121B1 DETECTOR STATE 1111 Tu 11:55:35 M19 N42121B1 DETECTOR STATE 1111 Tu 11:55:36 M19 N42121B1 DETECTOR STATE 1100 Tu 11:55:37 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:38 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:39 M19 N42121B1 DETECTOR STATE 0000 Tu 11:55:40 M19 N42121B1 DETECTOR STATE 1111 Vehicle 5 Tu 11:55:41 M19 N42121B1 DETECTOR STATE 1000 Tu 11:55:42 M19 N42121B1 DETECTOR STATE 0000 TrafiCam Tu 11:55:24 M19 N42121Z1 DETECTOR STATE 0011 Vehicle 1 Tu 11:55:25 M19 N42121Z1 DETECTOR STATE 1110 Tu 11:55:26 M19 N42121Z1 DETECTOR STATE 0000 Tu 11:55:27 M19 N42121Z1 DETECTOR STATE 0000 Tu 11:55:28 M19 N42121Z1 DETECTOR STATE 1111 Vehicle 2 Tu 11:55:29 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:30 M19 N42121Z1 DETECTOR STATE 1011 Vehicle 3 Tu 11:55:31 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:32 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:33 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:34 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:35 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:36 M19 N42121Z1 DETECTOR STATE 1100 Tu 11:55:37 M19 N42121Z1 DETECTOR STATE 0011 Vehicle 4 Tu 11:55:38 M19 N42121Z1 DETECTOR STATE 1111 Tu 11:55:39 M19 N42121Z1 DETECTOR STATE 1000 Tu 11:55:40 M19 N42121Z1 DETECTOR STATE 0000 Tu 11:55:41 M19 N42121Z1 DETECTOR STATE 1111 Vehicle 5 Tu 11:55:42 M19 N42121Z1 DETECTOR STATE 0000 Tu 11:55:43 M19 N42121Z1 DETECTOR STATE 0000 A further example below shows the fourth vehicle delayed for considerably longer in the TrafiCam detection zone than over the loop. It appears that the fifth vehicle arrived some time after the fourth had started queuing in the TrafiCam detection zone and queued over the loop until the fourth vehicle started to move. TRL 10 PPR434
SCOOT We 10:37:19 M19 N42121B1 DETECTOR STATE 0000 We 10:37:20 M19 N42121B1 DETECTOR STATE 0001 Vehicle 1 We 10:37:21 M19 N42121B1 DETECTOR STATE 1100 We 10:37:22 M19 N42121B1 DETECTOR STATE 0011 Vehicle 2 We 10:37:23 M19 N42121B1 DETECTOR STATE 1110 We 10:37:24 M19 N42121B1 DETECTOR STATE 0000 We 10:37:25 M19 N42121B1 DETECTOR STATE 0011 Vehicle 3 We 10:37:26 M19 N42121B1 DETECTOR STATE 1111 We 10:37:27 M19 N42121B1 DETECTOR STATE 0011 Vehicle 4 We 10:37:28 M19 N42121B1 DETECTOR STATE 1111 We 10:37:29 M19 N42121B1 DETECTOR STATE 1100 We 10:37:30 M19 N42121B1 DETECTOR STATE 0000 We 10:37:31 M19 N42121B1 DETECTOR STATE 0000 We 10:37:32 M19 N42121B1 DETECTOR STATE 0000 We 10:37:33 M19 N42121B1 DETECTOR STATE 0011 Vehicle 5 We 10:37:34 M19 N42121B1 DETECTOR STATE 1111 We 10:37:35 M19 N42121B1 DETECTOR STATE 1111 We 10:37:36 M19 N42121B1 DETECTOR STATE 1111 We 10:37:37 M19 N42121B1 DETECTOR STATE 0000 We 10:37:38 M19 N42121B1 DETECTOR STATE 0000 We 10:37:39 M19 N42121B1 DETECTOR STATE 0000 We 10:37:40 M19 N42121B1 DETECTOR STATE 0000 We 10:37:41 M19 N42121B1 DETECTOR STATE 0011 Vehicle 6 We 10:37:42 M19 N42121B1 DETECTOR STATE 1111 We 10:37:43 M19 N42121B1 DETECTOR STATE 0000 TrafiCam We 10:37:20 M19 N42121Z1 DETECTOR STATE 0000 We 10:37:21 M19 N42121Z1 DETECTOR STATE 0111 Vehicle 1 We 10:37:22 M19 N42121Z1 DETECTOR STATE 1000 We 10:37:23 M19 N42121Z1 DETECTOR STATE 0111 Vehicle 2 We 10:37:24 M19 N42121Z1 DETECTOR STATE 1110 We 10:37:25 M19 N42121Z1 DETECTOR STATE 0000 We 10:37:26 M19 N42121Z1 DETECTOR STATE 0001 Vehicle 3 We 10:37:27 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:28 M19 N42121Z1 DETECTOR STATE 1011 Vehicle 4 We 10:37:29 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:30 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:31 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:32 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:33 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:34 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:35 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:36 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:37 M19 N42121Z1 DETECTOR STATE 1000 We 10:37:38 M19 N42121Z1 DETECTOR STATE 0001 Vehicle 5 We 10:37:39 M19 N42121Z1 DETECTOR STATE 1111 We 10:37:40 M19 N42121Z1 DETECTOR STATE 1110 We 10:37:41 M19 N42121Z1 DETECTOR STATE 0000 We 10:37:42 M19 N42121Z1 DETECTOR STATE 0011 Vehicle 6 We 10:37:43 M19 N42121Z1 DETECTOR STATE 1110 We 10:37:44 M19 N42121Z1 DETECTOR STATE 0000 Other examples of the loop not staying on for longer than 3.5 seconds are not clear with a vehicle being detected for much longer by the TrafiCam than by the loop, but detected by both for part of the time. There may then be an appreciable gap on the loop, but that TRL 11 PPR434
could be that the original vehicle has moved forward off the loop, but still detected by the TrafiCam and it is the last vehicle in the queue. In some cases the loop recorded two vehicles with a gap between them of a second or so, but the TrafiCam only recorded one vehicle, but for a long time; subsequent vehicles were detected by both the loop and TrafiCam. TRL 12 PPR434
5 Conclusions The overall agreement between the TrafiCam and the inductive loop was good. The TrafiCam detection system tested for SCOOT seemed to work well in comparison with a loop system. Because of the need for the TrafiCam to see the rear lights of vehicles, its detection zone was not coincident with the inductive loop. Therefore, comparison of individual vehicle counts in slow moving traffic was more difficult than it would have been with coincident detection zones. Nevertheless, 99% of individual vehicle detections were matched. Total occupancy recorded by the TrafiCam was about 10% greater than the loop. Some of this difference could have been due to a slightly longer effective detection zone. In addition, the TrafiCam detection zone was downstream of the loop and, therefore, more likely to be queued in, resulting in more vehicles being delayed in its detection zone, hence greater occupancy, than over the loop. The SCOOT validation process would allow for a small difference in the length of the detection zone, which would then have no practical effect on the operation of SCOOT. There were some differences in the outputs of the two detectors when traffic was queuing back to the detector zones. Some of the differences could be explained in terms of the difference in detection zones, but in a very few cases it appeared that the TrafiCam might have stuck on for some time (up to half a minute) during which the loop detected several vehicles. Acknowledgements The work described in this report was carried out in the Traffic Group of the Transport Research Laboratory. The author is grateful to I J Burrow who carried out the technical review and auditing of this report. The research would not have been possible without the cooperation of Hampshire County Council and the author would like to acknowledge Adrian Gray for authorising the use of Hampshire facilities and Roger Derrick for his technical assistance and data collection. TRL 13 PPR434
Test of TrafiCam vehicle detector in Winchester A Traficam video vehicle detector was installed adjacent to a standard SCOOT inductive loop detector in Winchester. Outputs from the two detectors were returned to SCOOT and recorded for analysis. The inductive loop output was used in the SCOOT control algorithms as normal, but the trial Traficam detector was ignored by them. The status of both detectors, off or on, each quarter second was output in standard SCOOT messages and compared. The comparison was complicated as it was necessary to define the Traficam detection zone slightly downstream of that of the loop, but it was possible to make valid comparisons. Other titles from this subject area PPR206 International comparison of roundabout design guidelines. J Kennedy. 2008 PPR240 Behaviour at cycle Advanced Stop Lines. D Allen, S Bygrave and H Harper. 2007 PPR252 PPR275 Survey of MOVA and SCOOT operation at M42 Junction 6. K Wood, M Crabtree, A Kirkham, A Maxwell and R Robbins. 2008 Development of a human factors road safety assessment tool: interim report. D Basacik, T Luke and T Horberry. 2007 PPR292 A review of simplified Streetscape Schemes. A Quimby and J Castle. 2007 PPR298 The travel of errant vehicles after leaving the carriageway. D A Lynam and J V Kennedy. 2008 Price code: 2X ISSN 0968-4093 TRL Crowthorne House, Nine Mile Ride Wokingham, Berkshire RG40 3GA United Kingdom T: +44 (0) 1344 773131 F: +44 (0) 1344 770356 E: enquiries@trl.co.uk W: www.trl.co.uk Published by IHS Willoughby Road, Bracknell Berkshire RG12 8FB United Kingdom T: +44 (0) 1344 328038 F: +44 (0) 1344 328005 E: trl@ihs.com W: http://emeastore.ihs.com PPR434