Integrated Pedestrian Detection and Localization Using Stereo Cameras - Digital Signal Processing for In-Vehicle Systems and Safety - page 223

16.7 Conclusion

In this chapter, we proposed a method for pedestrian detection in traffic areas. We
integrate typical object detection method with sparse depth estimation. This enables
us to use 3D depth information naturally and improve detection accuracy by taking
into account the human knowledge that “things become smaller when they move
farther.”
The efficiency of our integration was shown in our experiment. Without adding
too much processing time, our method could improve the performance of our
baseline detection system to a significant level, even close to a state-of-the-art
detection system [ 1 ]. For the latter one, processing time for the detection with the
same image size will cost nearly five times. Besides efficiency, another thing that
we found out from the experiment is that the utilization of depth is independent of
image resolution and instance size. This leads to stable improvement over the
baseline system for all different kinds of scenes.
However, some issues still exist in the current system. First, the depth informa-
tion that we introduced is obtained in an explicit way. This will, in some level, make
the system sensitive against error in depth estimation. Secondly, our system is not
good in handling occlusion and therefore quite weak in some crowd scenes. In the
future work, we will mainly focus on robust depth estimation and occlusion
handling.

References

1. Felzenszwalb P, Girshick RB, McAllester D, Ramanan D (2010) Object detection with
discriminatively trained part based models. IEEE Trans Pattern Anal Mach Intell 32
(9):1627-1645
2. Torralba A, Murphy KP, Freeman WT (2004) Sharing features: efficient boosting procedures
for multiclass object detection. In: IEEE Conference on computer vision and pattern recogni-
tion, Washington, 2004
3. Leibe B, Schindler K, Cornelis N, Van Gool L (2008) Coupled object detection and tracking
from static cameras and moving vehicles. IEEE Trans Pattern Anal Mach Intell

30(10):1683-1698
4. Gavrila DM, Munder S (2007) Multi-cue pedestrian detection and tracking from a moving
vehicle. Int J Comput Vision 73(1):41-59
5. Hoiem D, Efros AA, Hebert M (2008) Putting objects in perspective. Int J Comput Vision 80
(1):3-15
6. Dalal N, Bill Triggs B (2005) Histograms of oriented gradients for human detection In:. IEEE
conference on computer vision and pattern recognition, San Diego, 2005
7. Joachims T (1999) Making large-scale support vector machine learning practical, MIT Press
Cambridge, MA, USA
8. Wang Y, Kato J (2010) Reference view generating of traffic intersection. ICIC Expr Lett
4(4):1083-1088
9. Ess A, Leibe B, Schindler K, Van Gool L (2009) Robust multi-person tracking from a mobile
platform. IEEE Trans Pattern Anal Mach Intell 31(10):1831-1846
10. Lowe DG (2008) Distinctive image features from scale-invariant keypoints. Int J Comput
Vision 60(2):91-110