Pedestrian Stride Frequency and Length Estimation in Outdoor Urban Environments using Video Sensors
RMSE for stride frequency (Hz) Rouen 0.170 (France) (0.123) Vancouver 0.161 (Canada) (0.090)
4
6
8
10
12
14
RMSE for the stride frequency
Power
0
1.4
4
6
8
10
12
14
Frequency (Hz) 40
1.2
1
2
20
0.8
0
0.6
0
50
100
150
200
250
4
6
8
10
12
14
Frequency (Hz)
0.4
0
2
10
300
Time (Frame)
5
0 0
2
4
6
8
10
12
14
Frequency (Hz)
Power Spectrum A set F of maximum Nf frequencies may satisfy conditions 1 and 2
Final stride frequency f = mean(F) or argmaxfF(Power(f))
Stride length l=d/(∆t×f)
g h i j
1 0.5 0
0
50
100
150 Frame
200
250
0.6
0.35
0.55
0.35
0.5
0.3
0.45 0.25
0.4 0.35
0.2
0.3 0.25
0.15
0.3
0.25
0.2
0.15
0.2 0.15
0.1 0.8
1
1.2
1.4
First frequency of the range [f
,f
min max
0.1 0.8
1.6
1
] (Hz)
1.2
1.4
First frequency of the range [f
,f
min max
Rouen dataset, Nf =10
1.6
] (Hz)
Vancouver dataset, Nf =20
Type predicted by the classification method Motorized vehicles Pedestrians Unknown 87 2 5 6 95 1
300
Experimental Results
Another Cue to Classify Pedestrians and Motorized Vehicles
700
250
Motorized vehicles 3
l < lpedestrian yes
no
yes
no
200
2.5
500
1.5
1
0.5
0
50
100
150
100
50
Time (Frame) 60
100
40
0 1.4
20
1.6
1.8
2
2.2
Stride frequency (Hz)
2.4
2.6
2.8
0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Stride length (m)
0 0
2
4
6
8
10
12
14
Frequency (Hz)
Power
Unknown
300
150
0
20
Motorized Vehicle
400
200
30
Pedestrian
Number of pedestrians
2
Number of pedestrians
a frequency f’ can be computed in [0, fmin] and noscillations < npedestrian oscillations
600
Speed (m/s)
A frequency f can be computed, and noscillations > npedestrian oscillations yes no
Observation: speed fluctuates at each stride
10
0 0
2
4
6
8
Frequency (Hz)
(noscillations: number of times per second that the speed profile goes through its mean value) Sample data from Automated Collection Of Pedestrian Data Using Computer Vision Techniques. Ismail, K., Sayed, T. and Saunier, N. TRB Annual Meeting , 2009
0.65
True type Motorized vehicles Pedestrians
2 1.5
0.4
Confusion Matrix for Road User Classification
time duration ∆t
distance d
0.4
0
RMSE for the stride frequency
Ratio of the maximum power
10
1.6
+Fast Fourier Transform (FFT) + Absolute Speed Time Series Value
Number of pedestrians with calculable stride frequency 101 / 102 (75) 42 / 50 (11)
Performance Evaluation
Frequency (Hz) 20
1.8
2
Ratio of the maximum power
2
0
mean frequency selection method
2.2
RMSE for stride length (m) 0.061 (0.040) 0.057 (0.030)
0
Power
Smoothing +Mean value subtraction
Pedestrians 2.4
20
Power
From Crowd-structure interaction in lively footbridges under synchronous lateral excitation: A literature review. Venuti, F. and Bruno, L. 3, 2009, Physics of Life Reviews, Vol. 6, pp. 176-206
Dataset
40
Motorized Vehicles
Relevant Work Biomechanics and transportation research Structural engineering: footbridge dynamic behavior under human loading (London Millennium Footbridge closed in 2000) Stride length and frequency are not commonly measured, even less automatically and non-intrusively in the field Walking parameter Range of the mean Range of the standard deviation Walking speed (m/s) 1.19 – 1.60 0.15 – 0.63 Stride frequency (Hz) 1.82 – 2.0 0.11 – 0.186 Stride length (m) 0.75 – 0.768 0.07 – 0.098
Condition 2 Condition 1 Threshold: Frequency Search Interval [fmin, fmax] ratio α of the maximum power
Power
Objective: extract automatically pedestrian stride frequency and length from video data collected non-intrusively in outdoor urban environments. Pedestrian walking gait is usually described by the relationship v=f×l , with the following walking parameters the walking velocity v the vertical stride frequency f (number of times a foot touches the ground per time unit) the stride length l
Experimental Validation
Proposed Method
Power
Introduction Walking is a key non-motorized mode of travel and a vital component of most trips Walking has traditionally received research and practice focus secondary to motorized modes There is a lack of pedestrian data, in particular microscopic data, to meet the analysis and modeling needs Distributions based on empirical measures are crucial for studies trying to estimate the impact of a shift from motorized modes to active transportation on the level of physical activity
Speed (m/s)
Annual Meeting 2011
N. Saunier1, A. El Husseini1, K. Ismail2, C. Morency1, J.-M. Auberlet3 and T. Sayed4 1École Polytechnique de Montréal, 2Carleton University, 3Université Paris Est, LEPSIS, IFSTTAR, 4University of British Columbia
Speed (m/s)
90th
10
12
14
Dataset
Stride frequency (Hz) Rouen annotated dataset (manual) 1.908 ± 0.214 1.901 ± 0.173 annotated dataset (auto) 1.897 ± 0.147 whole dataset Vancouver annotated dataset (manual) 1.703 ± 0.311 1.753 ± 0.174 annotated dataset (auto)
Stride length (m) 0.748 ± 0.139 0.759 ± 0.163 0.678 ± 0.217 0.625 ± 0.119 0.679 ± 0.132
Acknowledgements: the data was collected in the French National project SICAP from PISTES framework, sponsored by the Road Safety Foundation. The authors wish to thank David Doucet from the CETE-NC (Rouen, France) for the video collection, and Professor Christian Cardinal (École Polytechnique de Montréal) for his help with the Fourier analysis.
