A High Accuracy Pedestrian Detection System Combining a Cascade AdaBoost Detector and Random Vector Functional-Link Net
Table 1
Pedestrian detection rate (%) and false positives per frame comparison of CAB-ELM and other state-of-the-art detectors.
Data sets
ā
Propose (low)
Propose (high)
CAB
SVM
GAB
HF
Multiple Flow
PDR
94.55
94.72
94.39
94.22
61.55
78.99
FPPF
0.56
0.75
0.30
1.68
1.70
1.66
City Center
PDR
90.49
95.29
93.10
95.15
66.04
80.22
FPPF
0.46
1.80
0.71
0.960
1.43
1.50
USC
PDR
73.16
81.79
81.15
80.83
36.74
34.82
FPPF
0.24
0.88
0.34
0.91
0.34
0.42
INRIA
PDR
76.28
89.25
71.84
89.08
62.29
34.82
FPPF
0.59
9.24
3.54
9.55
0.89
0.42
(1) The number of cascade stages of CAB and CAB-RVFL is 12. (2) (low) and (high) means two sets with (low/high) detection rates and corresponding false positives per frame.