Research Project MOBIS: Behaviour of Road Users at Behaviour of Road Users at „Development of a method for assessing safety of Unsignali Unsignaliz zed Pedestrian Crossings ed Pedestrian Crossings pedestrian crossings using automatic video analysis” in Poland in Poland � Consortium: Warsaw University of Technology, Motor Transport Institute, Neurosoft Ltd. � Financing by Polish „National Centre for Research and Development” ( Applied Research Programme) � Time frame: 3 years 2012 – 2015 Piotr Olszewski, Witold Czajewski, Paweł D ą bkowski, Piotr Szagała Warsaw University of Technology Warsaw University of Technology Ilona Buttler Motor Transport Institute Motor Transport Institute 28 th ICTCT Workshop, Ashdod, 29-30 October 2015 28th ICTCT Workshop, Ashdod 29-30 October 2015 2 2 Problem significance Field surveys and tests � Aim: development of a method for assessing the � Pedestrian road safety situation in Poland (2014): safety of pedestrians using automatic video analysis – 1116 pedestrians killed (35% of all traffic fatalities) � Basis: identification of dangerous encounters (traffic – 8398 pedestrians injured (20% of all traffic injuries) conflicts - events which could lead to an accident) � During 6 years (2008-2013): between vehicles and pedestrians – 13% pedestrians were killed and � Assessments based on surrogate measures can – 26% were injured at unsignalized zebra crossings hopefully use relatively short observation periods � Road safety situation is � During the project, four field tests were conducted generally improving at different crossings, using different safety � Accidents at pedestrian improvement measures crossings have not decreased in the last 4 years 28th ICTCT Workshop, Ashdod 29-30 October 2015 3 3 28th ICTCT Workshop, Ashdod 29-30 October 2015 4 4 Test site POW in Warsaw Test sites in Wrocław (CEN, SWO) � 2 zebra crossings near tram stop � Zebra crossing with pedestrian � 2-lane 2-way road refuge in the middle � Installation of system Levelite – � 4-lane 2-way road � Temporary installation of LED lights embedded in pavement � CEN direction: continuous blinking SignFlash system – yellow � SWO direction: blinking after lights flashing after detecting pedestrian detection a pedestrian 28th ICTCT Workshop, Ashdod 29-30 October 2015 5 5 28th ICTCT Workshop, Ashdod 29-30 October 2015 6 6
Data collection Video data analysis � Video recording system: an overview � Motion trajectories of vehicles and pedestrians were camera plus one directional camera per determined based on video processing. lane, terminal � Several parameters describing pedestrian-vehicle � Only days with good conditions (no shadows, no traffic jam) were selected encounters were calculated, such as: for processing � Speed profile of pedestrians at the zebra � Speed profile of vehicles (30 m approach to zebra) Period No of days No of days Safety � Vehicle deceleration (average) Site with video selected for measure recording analyses used from to � Minimum distance between event participants Warsaw SignFlash (SF) 23.09.2013 19.12.2013 49 23 � PET value POW Wroc ł aw 01.08.2014 27.11.2014 103 18 LeveLite (LL) CEN 28th ICTCT Workshop, Ashdod 29-30 October 2015 7 7 28th ICTCT Workshop, Ashdod 29-30 October 2015 8 8 Classification of vehicle-pedestrian encounters Classification of vehicle-pedestrian encounters � A1 - vehicle passes � B - vehicle passes directly in front of a immediately behind a pedestrian who is on the pedestrian who is on the zebra crossing; zebra crossing; � A2 - vehicle passes directly in front of a � C - vehicle slows down or pedestrian who is on the stops before the crossing. sidewalk; 28th ICTCT Workshop, Ashdod 29-30 October 2015 9 9 28th ICTCT Workshop, Ashdod 29-30 October 2015 10 10 Classification of vehicle-pedestrian encounters Speed of vehicles and pedestrians Yielding � Mean speed profiles of vehicles over the distance of 30 m before the crossing were analysed Site Safety Encounters A1 A2 B C Sum � Speed profiles for encounters A1 and C: without SF 7088 3,9% 16,2% 15,4% 64,5% 100% Warsaw POW (both lanes) active SF 6418 4,6% 15,9% 14,5% 65,0% 100% without LL 11519 11,3% 41,7% 8.4% 38.6% 100% Wrocław CEN steady LL 3197 11.9% 33.8% 9.2% 45.1% 100% without LL 4425 15.1% 40.8% 7.2% 36.9% 100% Wrocław SWO active LL 3289 13.7% 33.9% 9.4% 43.0% 100% 28th ICTCT Workshop, Ashdod 29-30 October 2015 11 11 28th ICTCT Workshop, Ashdod 29-30 October 2015 12 12
Effect of safety measures on approach speed Speed of vehicles and pedestrians Warsaw POW site Situa- Time of Diffe- � Speed profiles for encounters B and A2: Without SF With SF Sig. level tion day rence n V* [km/h] n V* [km/h] [km/h] Day 197 45.0 237 41.4 -3.6 0.01 A1 Night 81 41.6 50 38.8 -2.7 0.10 Day 2647 15.0 3242 15.1 -0.1 no C Night 1909 15.5 917 15.7 -0.2 0.05 Wrocław CEN site Situa- Time of Diffe- Without LL With LL Sig. level tion day rence n V* [km/h] n V* [km/h] [km/h] Day 1079 39.2 323 33.5 -5.8 0.01 � Generally lower speeds in Wrocław (higher flow) A1 Night 222 41.8 56 32.7 -9.0 0.01 � Levelite more effective than SignFlash Day 3763 19.0 1161 16.9 -2.1 0.01 C Night 680 20.6 282 16.8 -3.8 0.01 * Speed measured 10 m from the zebra crossing 28th ICTCT Workshop, Ashdod 29-30 October 2015 13 13 28th ICTCT Workshop, Ashdod 29-30 October 2015 14 14 Pedestrian speed distribution Survey of pedestrian risk perception 140 100% � Video clips of vehicle-pedestrian encounters which without vehicles 120 with vehicles seemed dangerous were extracted and used in a 80% cum. dist. without vehicles 100 cum. dist. with vehicles survey of pedestrian risk perception. Pedestrian count 60% 80 � Viewers were asked to rate the situations on a scale 60 from 1 to 10 from the point of view of pedestrians 40% 40 � 1 = very safe, no risk 20% 20 to pedestrian 0 0% � 10 = very dangerous, 0,70 0,75 0,80 0,85 0,90 0,95 1,00 1,05 1,10 1,15 1,20 1,25 1,30 1,35 1,40 1,45 1,50 1,55 1,60 1,65 1,70 1,75 1,80 1,85 1,90 1,95 2,00 2,05 2,10 2,15 2,20 2,25 2,30 near accident Pedestrian velocity [m/s] � Viewers were students � Pedestrians tend to walk faster by ~2% (0.04 m/s) and traffic safety experts when vehicles are approaching 28th ICTCT Workshop, Ashdod 29-30 October 2015 15 15 28th ICTCT Workshop, Ashdod 29-30 October 2015 16 16 Encounter Risk Indicator Distribution of calculated ERI � Formula calibrated by non-linear regression for encounters A1 and B: V v – vehicle speed at min. distance (m/s) S – minimum distance between vehicle and pedestrian (m) X – vehicle passes: 0 = in front of, 1 = behind a pedestrian � Problem: the formula is not very sensitive to speed � Calculated W show no positive effect of safety measures: � Percentage of encounters with ERI W > 3.0: SF and LL Warsaw POW = 15.0%, Wrocław CEN = 10.6% 28th ICTCT Workshop, Ashdod 29-30 October 2015 17 17 28th ICTCT Workshop, Ashdod 29-30 October 2015 18 18
Distributions of observed minimum distance Conclusions � Parameters which could be extracted from video: speed of pedestrians and vehicles, deceleration, minimum distance between users, PET value � Classification of encounters is based on who passes first and on relative position on the zebra � Levelite system considerably increases the percentage of yielding � Levelite also causes a significant decrease of the approaching vehicle speed (at 10 m) – between 2.1 and 9.0 km/h � Percentage of encounters with S ≤ 1.5 m: Warsaw = 3.9%, Wrocław = 2.2% 28th ICTCT Workshop, Ashdod 29-30 October 2015 19 19 28th ICTCT Workshop, Ashdod 29-30 October 2015 20 20 Conclusions � User perception survey was conducted to rate the perceived danger of pedestrian-vehicle encounters � The average risk scores correlate well with the following variables: � Minimum distance between vehicle and pedestrian � Vehicle speed at minimum distance � Type of encounter � Maximum deceleration � Distribution of minimum distance can be used to identify the most safety-critical encounters 28th ICTCT Workshop, Ashdod 29-30 October 2015 21 21
Recommend
More recommend
